Occasional Paper No. 52
A safe and supportive family environment
for children: key components and links to
child outcomes
Killian Mullan and Daryl Higgins
Australian Institute of Family Studies
© Commonwealth of Australia 2014
ISSN 1839-2334
ISBN 978-1-925007-50-3
All material presented in this publication is provided under a Creative Commons CC-BY Attribution 3.0
Australia licence.
For the avoidance of doubt, this means this licence only applies to material as set out in this document.
With the exception of the Commonwealth Coat of Arms (refer to terms of use), the details of the relevant licence
conditions are available on the Creative Commons website, as is the full legal code for the CC-BY 3.0 AU
licence.
The views expressed in this publication are those of individual authors and may not reflect those of the
Australian Government or the Australian Institute of Family Studies and cannot be taken in any way as
expressions of government policy.
Administrative Arrangements Orders changes
On 18 September 2013 the Department of Families, Housing, Community Services and Indigenous Affairs
(FaHCSIA) was renamed the Department of Social Services (DSS). References in this publication to FaHCSIA
should be read in that context.
Acknowledgements
This research paper makes use of data from Growing Up in Australia: the Longitudinal Study of Australian
Children (LSAC). LSAC is conducted in partnership between the Department of Social Services (formerly the
Department of Families, Housing, Community Services and Indigenous Affairs [FaHCSIA]), the Australian
Institute of Family Studies (AIFS) and Australian Bureau of Statistics (ABS), with advice provided by a
consortium of leading researchers.
The authors would like to thank colleagues at AIFS and DSS for their valuable comments on previous versions of
this report.
At the time of writing, Dr Killian Mullan was a Research Fellow at AIFS (he is currently employed at the
London School of Economics). Dr Daryl Higgins is the Deputy Director (Research) at AIFS.
For more information contact
Research Publications Unit
Policy Strategy Branch
Australian Government Department of Social Services
PO Box 7576
Canberra Business Centre ACT 2610
Phone: (02) 6146 8061
Fax: (02) 6293 3289
Email: publications.research@dss.gov.au
CONTENTS
Contents
EXECUTIVE SUMMARY ..................................................................................................... IX
1
INTRODUCTION.......................................................................................................... 1
2
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS ....................................................... 3
2.1
Child safety and wellbeing—the Australian policy context ..................................................................3
2.2
Public health approach to child protection: implications for research ...........................................6
3
CLASSIFYING FAMILY ENVIRONMENTS: METHODOLOGY........................................... 12
3.1
Latent class cluster analysis .......................................................................................................................... 12
3.2
Data and sample ................................................................................................................................................ 12
3.3
Measures............................................................................................................................................................... 14
4
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS ...................................................... 16
4.1
Identifying different family environments .............................................................................................. 16
4.2
Latent class profiles .......................................................................................................................................... 18
4.3
Characteristics of families in each latent class ....................................................................................... 27
4.4
Summary............................................................................................................................................................... 30
5
FAMILY ENVIRONMENTS AND CHILD OUTCOMES ..................................................... 32
5.1
Children’s health ................................................................................................................................................ 33
5.2
Children’s social and emotional wellbeing .............................................................................................. 41
5.3
Children’s cognitive development .............................................................................................................. 48
5.4
Summary............................................................................................................................................................... 55
6
TRANSITIONS IN THE FAMILY ENVIRONMENT AND CHILD OUTCOMES ...................... 57
6.1
A descriptive overview of transitions in family environments ........................................................ 57
6.2
Multivariate analysis of transitions and child outcomes ................................................................... 61
6.3
Summary............................................................................................................................................................... 63
7
7.1
SUMMARY AND DISCUSSION ................................................................................... 65
Background ......................................................................................................................................................... 65
iii
CONTENTS
7.2
Aims of the research ......................................................................................................................................... 65
7.3
Analysis ................................................................................................................................................................. 65
7.4
Findings: identifying family environments ............................................................................................. 66
7.5
Findings: family environments and child outcomes ............................................................................ 68
7.6
Implications for policy..................................................................................................................................... 69
7.7
Limitations ........................................................................................................................................................... 71
7.8
Conclusion ............................................................................................................................................................ 73
APPENDIX: TECHNICAL DETAILS ABOUT CLASS SELECTION ............................................... 75
LIST OF SHORTENED FORMS............................................................................................ 77
REFERENCES ................................................................................................................... 78
ENDNOTES ...................................................................................................................... 83
iv
CONTENTS
List of tables
Table 1:
Child protection notifications in Australia, children 0–17 years, 1989–90 to 2009–10
4
Table 2:
Final class counts and proportions: families with two resident parents .................. 17
Table 3:
Final class counts and proportions: families with a PLE ........................................... 17
Table 4:
Relative risk ratios from multinomial logistic regression predicting membership of
disengaged or enmeshed relative to cohesive: families with two resident parents ............ 29
Table 5:
Relative risk ratios from multinomial logistic regression predicting membership of
disengaged or enmeshed relative to cohesive families: families with a PLE ....................... 30
Table 6:
Relative risk ratios from multinomial logistic regression predicting children’s
weight problems: families with two resident parents .............................................................. 36
Table 7:
Relative risk ratios from multinomial logistic regression predicting children’s
weight problems: families with a PLE ....................................................................................... 37
Table 8:
Odds ratios from logistic regression predicting children’s injury (one or more in
past 12 months): families with two resident parents ............................................................... 40
Table 9:
Odds ratios from logistic regression predicting children’s injury (one or more times
in past 12 months): families with a PLE .................................................................................... 41
Table 10: Children’s SDQ prosocial score: families with two resident parents (OLS
coefficients).................................................................................................................................. 43
Table 11:
Children’s SDQ prosocial score: families with a PLE (OLS coefficients) ................. 44
Table 12: OLS coefficients predicting children’s SDQ total problem score: families with two
resident parents ........................................................................................................................... 47
Table 13:
OLS coefficients predicting children’s SDQ total problem score: families with a PLE
48
Table 14: OLS coefficients predicting children’s NAPLAN reading score: families with two
resident parents ........................................................................................................................... 51
Table 15:
OLS coefficients predicting children’s NAPLAN reading score: families with a PLE
52
Table 16: OLS coefficients predicting children’s NAPLAN numeracy score: families with two
resident parents ........................................................................................................................... 54
Table 17:
PLE
OLS coefficients predicting children’s NAPLAN numeracy score: families with a
55
Table 18: Cross-tabulation of family environments: B cohort children 2–3 years and 6–7
years, families with two resident parents ................................................................................. 58
Table 19: Cross-tabulation of family environments: K cohort children 4–5 years and 10–11
years, families with two resident parents ................................................................................. 58
Table 20: Cross-tabulation of family environments: B cohort children 2–3 years and 4–5
years, families with a PLE ........................................................................................................... 59
v
CONTENTS
Table 21: Cross-tabulation of family environments: K cohort children 6–7 years and 10–11
years, families with a PLE ........................................................................................................... 60
Table 22: Relative risk ratios from multinomial logistic regression predicting transition:
B cohort children 2–3 years and 6–7 years .............................................................................. 60
Table 23: Relative risk ratios from multinomial logistic regression predicting transition:
K cohort children 4–5 years and 10–11 years .......................................................................... 61
Table 24: Coefficients from two-wave panel regression of indicators of children’s social and
emotional wellbeing, and cognitive development: K cohort children 4–5 years and 10–11
years 62
Table 25: Coefficients from two-wave panel regression of indicators of children’s social and
emotional wellbeing, and cognitive development: K cohort children 6–7 years and 10–11
years: families with a PLE .......................................................................................................... 63
Appendix Tables
Table A1:
Sample-adjusted BIC and Vuong-Lo-Mendell-Rubin p-values: families with two
resident parents
Table A2:
Sample-adjusted BIC and Vuong-Lo-Mendell-Rubin p-values: families with PLE
vi
CONTENTS
List of figures
Figure 1: Wave selection for families with two resident parents ................................................... 13
Figure 2: Wave selection for families with a parent living elsewhere (PLE) ................................. 13
Figure 3: Average standardised factor scores for parental warmth, angry parenting, parental
conflict and parent–child shared activities in disengaged (Class 3), cohesive (Class 1) and
enmeshed (Class 2) families: B cohort 2–3 years .................................................................... 20
Figure 4: Average standardised factor scores for parental warmth, angry parenting, parental
conflict and parent–child shared activities in disengaged (Class 3), cohesive (Class 1) and
enmeshed (Class 2) families: K cohort 4–5 years .................................................................... 21
Figure 5: Average standardised factor scores for parental warmth, angry parenting, parental
conflict and parent–child shared activities in disengaged (Class 1), cohesive (Class 2) and
enmeshed (Class 3) families: B cohort 6–7 years .................................................................... 22
Figure 6: Average standardised factor scores for parental warmth, angry parenting, parental
conflict and parent–child shared activities in disengaged (Class 1), cohesive (Class 2) and
enmeshed (Class 3) families: K cohort 10–11 years ................................................................ 23
Figure 7: Average standardised factor scores for parental warmth, parental conflict and
parent–child shared activities in disengaged (Class 1) and cohesive (Class 2) families with
a PLE: B cohort 2–3 years .......................................................................................................... 24
Figure 8: Average standardised factor scores for parental warmth, parental conflict and
parent–child shared activities in disengaged (Class 2), cohesive (Class 1) and enmeshed
(Class 3) families with a PLE: B cohort 4–5 years ................................................................... 25
Figure 9: Average standardised factor scores for parental warmth, parental conflict and
parent–child shared activities in disengaged (Class 1), cohesive (Class 3) and enmeshed
(Class 2) families with a PLE: K cohort 6–7 years ................................................................... 26
Figure 10: Average standardised factor scores for parental warmth, parental conflict and
parent–child shared activities in disengaged (Class 2), cohesive (Class 1) and enmeshed
(Class 3) families with a PLE: K cohort 10–11 years ............................................................... 27
Figure 11: Average percentage of children underweight or overweight/obese in different family
groups: children 2–3 years, 4–5 years, 6–7 years and 10–11 years: families with two
resident parents ........................................................................................................................... 35
Figure 12: Average percentage of children underweight or overweight/obese in different family
groups: children 2–3 years, 4–5 years, 6–7 years and 10–11 years: families with a PLE .... 35
Figure 13: Average percentage of children with one or more injuries in the past year in different
family groups: children 2–3 years, 4–5 years, 6–7 years and 10–11 years: families with two
resident parents ........................................................................................................................... 38
Figure 14: Average percentage of children with one or more injuries in the past year in different
family groups: children 2–3 years, 6–7 years and 10–11 years: families with a PLE ........... 39
Figure 15: Average SDQ prosocial score in different family groups: children 4–5 years, 6–7
years and 10–11 years: families with two resident parents .................................................... 42
Figure 16: Average SDQ prosocial score in different family groups: children 4–5 years, 6–7
years and 10–11 years: families with a PLE ............................................................................. 42
vii
CONTENTS
Figure 17: Average SDQ total problem score in different family groups: children 4–5 years, 6–7
years and 10–11 years: families with two resident parents .................................................... 45
Figure 18: Average SDQ total problem score in different family groups: children 4–5 years, 6–7
years and 10–11 years: families with a PLE ............................................................................. 46
Figure 19: Average Year 3 and Year 5 NAPLAN reading scores in different family groups:
children 8.5 years and 10.5 years (average age): families with two resident parents.......... 49
Figure 20: Average Year 3 and Year 5 NAPLAN reading scores in different family groups:
children 8.5 years and 10.5 years (average age): families with a PLE ................................... 49
Figure 21: Average Year 3 and Year 5 NAPLAN numeracy scores in different family groups:
children 8.5 years and 10.5 years (average age): families with two resident parents.......... 53
Figure 22: Average Year 3 and Year 5 NAPLAN numeracy scores in different family groups:
children 8.5 years and 10.5 years (average age): families with a PLE ................................... 53
viii
EXECUTIVE SUMMARY
Executive summary
Introduction
Families are the mainstay of safety and support for children. While most children live in safe
and supportive environments, governments are aware that too many children are becoming
known to child protection services. This has led to a shift in thinking away from solely
concentrating on responding to ‘risk of harm’ reports towards a broader public health
approach to protecting all of Australia’s children, reducing the likelihood of children coming
to the attention of statutory authorities. This report aims to understand more about the
prevalence of different types of family environments in society and to explore the influence of
these environments on different child outcomes. Using data from the Longitudinal Study of
Australian Children, we used latent class cluster analysis to identify different family
environments and analysed the associations between these environments and particular child
outcomes.
Key findings
Classifying the family environment: profiles and characteristics
We identified three broad groups embodying three types of family environments that were
closely aligned with previous theoretic and empirical research. In line with previous research,
we referred to these groups as:

cohesive: the largest group of families exhibited average or above-average levels of
parental warmth and parent–child shared activities, and below-average levels of
hostile parenting and parental relationship conflict

disengaged: a smaller group of families exhibited below-average levels of parental
warmth and parent–child shared activities, and above-average levels of hostile
parenting

enmeshed: a final, relatively small, group exhibited average levels of parental
warmth, but higher than average levels of conflict in the relationship between parents.
It is important to note that:

These are not absolute distinctions, but rather relative positions on a spectrum ranging
from highly disengaged to highly enmeshed. No family environment can be
comprehensively understood using a finite set of factors. In addition, we show that
family environments can and do change over time.

The vast majority of families sit around the middle; some families tend toward either
relatively more disengagement or enmeshment, but do not approach, or even come
close to, more extreme aspects of these types. However, the most problematic families
with respect to child protection will most likely be located toward the extremes of the
range (see Section 2.2).
Associations between family environment and child outcomes
Unless specified otherwise, results relate to children aged 2–3, 4–5, 6–7 and 10–11 years
growing up either in families with two adults residing together (whether married or
cohabiting) who both have day-to-day responsibility for the child (including biological,
adoptive, step, foster, and grandparents) or families where a parent lives elsewhere from
ix
EXECUTIVE SUMMARY
the child's primary carer. All comparisons are made in relation to families that were relatively
more cohesive.
There were not many significant associations between family environment (as measured in
this report) and health outcomes. Significant associations were restricted to children aged 2–
3 years with two resident parents. That is:

children of this age in families tending toward enmeshment were more likely to be
underweight (than normal weight)

children of this age in families that were relatively more disengaged were more likely to
have one or more injuries per year.
Family environments were very strongly associated with children’s social and emotional
wellbeing. That is:

children in families indicating disengagement had significantly lower levels of prosocial
behaviour and higher levels of problem behaviour

children in families indicating enmeshment had significantly lower levels of prosocial
behaviour and higher levels of problem behaviour (this was not significant for children 4–
5 and 10–11 years old in families with a parent living elsewhere).
There were less consistent and fewer significant associations between family environment and
children’s cognitive development. In families with two resident parents:

children in families scoring relatively high on disengagement averaged lower Year 5
NAPLAN1 reading and numeracy scores.
Associations between changes in family environment and changes in child
outcomes

Children in families with two resident parents whose family environment improved
(became more cohesive) showed improved social and emotional wellbeing; children
whose family environments became relatively more problematic exhibited increased
social and emotional problems.

In families with two resident parents, children gained higher NAPLAN reading scores if
their family environment became relatively more cohesive.
Policy implications
These findings suggest policy may be more effective if it:



is attuned or sensitive to different family environments
targets behaviours rather than groups of people
recognises that families can both change for the better, and draw on their own prior
(positive) experiences.
Finally, results linking family environments to key child outcomes (especially around social
and emotional wellbeing) provide a clear impetus for a public health approach promoting safe
and supportive family environments. These research findings may provide insights to support
different types of responses, including parenting programs, public information campaigns and
more targeted referrals for intensive family support.
x
1 Introduction
Because of their developmental needs, children are a vulnerable group. The family exists as
one of the most effective means of protecting them. Although the vast majority of families
provide a safe and supportive environment, they sometimes fail to protect children and—in
the worst instances—can cause them substantial harm. Unfortunately, there is some evidence
to suggest that such instances may be increasing, or at least that services are becoming more
aware of families at risk. This trend has led to a radical reform of the policy approach to child
protection. Rather than focusing solely on the primary or more severe manifestations of the
problem, scholars and policymakers have sought to adopt a broader public health approach to
child protection (see Section 2.1).
To reiterate, the vast majority of families provide a safe and supportive environment for
children, and a public health perspective does not seek to cast aspersions on families in
general. Rather, it seeks to draw on their strengths and learn from their difficulties in order to
ensure that all children thrive. As part of a public health approach to child protection,
governments are committed to monitoring a broad range of indicators of children’s welfare,
and to ensuring that all children live in safe and supportive families.
What can research contribute to this? First, as most research attention in this area focuses on
the most problematic families, we know relatively little about different family environments
in the broader population. There are many unanswered questions. For example, to what extent
do Australian children live in safe and supportive family environments? What other types of
family environments do children in the broader population live in? These questions reflect
gaps in our knowledge, and, if we are to improve the likelihood of children living in safe and
supportive environments, we must bridge these gaps. Second, research can provide important
insights into how children are faring on a range of indicators of child welfare. In the absence
of a comprehensive census of child outcomes, we must rely on survey data, and research has
an important role to play in scrutinising these data. Third, understanding more about the links
between the family environment and child outcomes may serve to underline the importance of
a safe and supportive family environment, providing further evidential support for a broader
public health approach to child protection.
This study has three interlinked objectives:



first, and central to this report, to identify different family environments in the broader
society that vary in the extent to which they exhibit features concordant with a safe
and supportive environment for children
second, to understand more about how Australian children are faring on a broad range
of wellbeing indicators
third, to understand more about the influence of different family environments on
specific child outcomes.
Chapter 2 outlines the background relating to the child protection policy setting and considers
previous theoretical and empirical work examining different family environments that draw
out the key research questions the report addresses. The next two chapters address the report’s
first key objective: Chapter 3 details the methodology we employ to identify different family
environments and Chapter 4 presents the results of this analysis, including a detailed
description of the different family environments we identify. The following two chapters
address the second and third key objectives of this study: concurrent associations between
child outcomes and different family environments are analysed in Chapter 5, and associations
between transitions in the family environment and changes in child outcomes in Chapter 6.
1
Chapter 7 contains a discussion of the results, along with some conclusions and implications
for policy and service provision.
2
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS
2 Safe and supportive family environments
Families are critically important for children’s development (Bowes, Watson & Pearson
2009). To grow up safe and well, children need the security and support of a well-functioning
family. Understanding the role that families play in creating safe and supportive environments
for children is important. Although families can be the source of harm (for example, from
child abuse, neglect or exposure to domestic violence), they can also be the most important
source of protection from harm for children when they provide a sense of security, foster selfesteem and respond appropriately to children’s needs.
Parents vary in the degree to which they use positive parenting behaviours. Some families
struggle to provide consistently warm, nurturing and safe environments. In the absence of
such safety and security, children need protection—and statutory systems provide the safety
net for responding to children in need of interventions to ensure their safety. In this chapter,
we outline the policy context for ensuring children’s safety, and the frameworks within which
governments and agencies in Australia operate to support vulnerable families and address
children’s wellbeing needs.
We also provide an overview of the child protection policy setting, including trends and
policy developments culminating in a public health approach to child protection (Section 2.1).
We then consider theoretical and empirical research on the family that provides important
background to our knowledge about the key components of a safe and supportive family
environment, as well as the features of other, more problematic, family environments that
present potential risks for children (Section 2.2). The chapter concludes by outlining our
research focus and setting out the specific research questions addressed in this report.
2.1
Child safety and wellbeing—the Australian policy context
Australian state, territory and Commonwealth governments all recognise the problem of child
abuse and family violence, and the importance not only of responding to the needs of victims
but also of early intervention and prevention. This includes a focus on domestic and family
violence, child maltreatment, bullying, elder abuse, and sexual assault and violence. Domestic
and family violence is recognised as a major social issue in Australia. Reviews of family law,
child protection services and the juvenile justice system reveal a common set of family
problems that typically lead to engagement with these service systems—that is, family
violence, mental health issues and addictions to alcohol, tobacco, other drugs and gambling
(Higgins & Katz 2008). The common feature of such parental behaviours or circumstances is
that they can impair a family’s capacity to provide positive parenting and ensure that children
are safe and protected from harm.
Scale of the problem
Although there is debate about whether the underlying incidence of maltreatment has
changed, there is no doubt that over the past two decades there has been a very large increase
in notifications to statutory child protection authorities (that is, reports of concerns relating to
the abuse or neglect of children). This increase can be seen in both the absolute number of
notifications (increasing the workload of the departmental staff who need to screen and
potentially respond to them) and the rate per thousand of children in the population who are
the subject of notifications (see Table 1).
Some children in need of protection need to be taken into the care of the state because their
family environment is so unsafe that their wellbeing would be seriously compromised if they
remained living with their parent or parents—this is one end of the spectrum. In line with the
3
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS
increase in notifications over the past two decades, there has also been a substantial increase
in the number of children residing in out-of-home care2, from 3.0 per thousand in 1990 to 7.0
per thousand in 2010 (Higgins 2011).
Table 1:
Year
1989–90
1999–2000
2009–10
Child protection notifications in Australia, children 0–17 years, 1989–90 to 2009–10
Notifications
42,695
107,134
286,437
Total population of children
4,188,795
4,766,920
5,092,806
Rate per 1,000
10.4
22.5
56.2
Sources: Australian Bureau of Statistics (2010); Australian Institute of Health and Welfare (2001, 2011a); WELSTAT
(1991). Reproduced from Higgins (2011).
Response frameworks
To respond to the challenge of child abuse and children’s exposure to family violence, two
significant frameworks or plans have been developed in Australia that set the direction for
addressing the causes and consequences of violence and maltreatment. They are
 the National Framework for Protecting Australia’s Children 2009–2020 (Council of
Australian Governments (COAG) 2009a)
 The National Plan to Reduce Violence Against Women and their Children 2010–2022
(COAG 2009b).
Both take a population-based or public health approach, focusing on the causes (also referred
to as risk factors or social determinants) of violence, abuse and neglect in order to reduce their
incidence. The application of the tenets of a public health approach to child safety and
wellbeing is based on the notion that actions are delivered across three tiers of prevention:
primary (universal supports for all families), secondary (support for families needing extra
assistance, with a focus on early intervention) and tertiary (child protection services). For
further information on how a public health approach is applied to child safety and the
protection of children, see Bromfield, Arney and Higgins (2014), Hunter (2011), O’Donnell,
Scott and Stanley (2008), or Scott, Higgins and Franklin (2012).
The National Framework for Protecting Australia’s Children has six key supporting
outcomes:
 children live in safe and supportive families and communities
 children and families access adequate support to promote safety and intervene early
 risk factors for child abuse and neglect are addressed
 children who have been abused or neglected receive the support and care they need for
their safety and wellbeing
 Indigenous children are supported and safe in their families and communities
 child sexual abuse and exploitation is prevented, and survivors receive adequate
support.3
Although community violence and assaults are part of their focus, these frameworks give
prominence to the problem of family violence and maltreatment and the strategies needed for
prevention, for early intervention in at-risk families and for responding to the needs of
victims.
Increasingly, researchers are recognising not only that children are often subjected to multiple
forms of abuse and neglect but also that they experience a range of other threats to their safety
and wellbeing, such as bullying and peer victimisation (Higgins 2010b). Prevention efforts are
being implemented in settings such as schools, community organisations, workplaces and the
media. There are also programs targeted at men and boys that aim to develop and promote
4
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS
gender-equal, respectful relationships with girls and women (Australian Institute of Family
Studies 2012).
Families remain the central focus—in terms of both the risks of maltreatment of children
(which are often characteristics or behaviours of parents) and protecting children. Although
families are not always the only site of violence and maltreatment of children, they can still—
along with other agencies and institutions—be enlisted to assist with interventions to support
children and keep them safe.
The National Framework for Protecting Australia’s Children articulated a shift away from
seeing the protection of children solely as the role of statutory authorities. From a public
health perspective, protecting children and creating safe environments for them starts at home.
However, it is not sufficient to simply ‘bolt on’ preventive programs to the current child
protection processes. Researchers and commentators have argued that the role and function of
child protection systems need to be reviewed in the context of the wider range of policies and
programs aimed at promoting the wellbeing of children. This is of particular importance in
Indigenous communities, for two reasons: firstly, the over-representation of Indigenous
children in statutory child protection activities and, secondly, the potential for communityowned and community-led initiatives to support the health, wellbeing and safety of
Indigenous children in culturally appropriate ways (Higgins & Katz 2008).
Policy settings to address children’s safety needs
In the next subsections, we briefly introduce four contexts in Australia that reflect the
importance of policies that aim to create or sustain the family dynamics and circumstances
that provide safe and supportive environments for children.
Family law and protection from violence
Parental separation can be a time of increased risk to the safety and wellbeing of children.
Family courts often face difficult choices when parents raise concerns about child abuse or
violence by their partner during disputes over children’s matters (Higgins 2007). In 2006,
Australia’s family law system underwent some significant changes designed to encourage
greater involvement by both parents in their children's lives after separation. These changes
also brought with them the responsibility for systems to protect children from violence and
abuse and ensure the safety of all family members. Further changes to the Family Law Act in
2012 strengthened the framework for responding to family violence and child abuse concerns
(Croucher 2014).
Alternative family policy: out-of-home care
In order to ensure the safety of children and young people, all states and territories in
Australia have a responsibility to monitor concerns about the risk of harm to children from
abuse and neglect. If protective measures cannot be put in place to ensure the safety of
children while in the care of their parents, statutory child protection departments have the
responsibility to intervene. The number of Australian children in alternative care (out-ofhome care) has grown considerably over the last two decades (see Table 1; Higgins 2011).
There is a strong push for children to be restored to—or maintain active contact with—their
parents; however, for reunification to be successful, the dysfunctional dynamics and
problematic parenting behaviours that typically bring families to the attention of child
protection authorities—including family violence, mental health and addictions—need to be
addressed.
Poverty
Although child abuse and neglect (particularly child sexual abuse) occur across all family
5
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS
forms and socioeconomic strata and are under-reported, poverty and social disadvantage are
generally associated with higher risks of harm, particularly from neglect (Higgins 2010a).
Key issues relating to the economic security of families are the availability and adequacy of
employment, and systems to support families on low incomes or experiencing unemployment
(Adema 2012; Australian Institute of Family Studies 2012; Howe 2012). Although Australia
has a relatively low level of joblessness overall, the number of Australian families in which
no adult member of the household is in paid employment is high compared to many other
Organisation for Economic Co-operation and Development (OECD) countries. This is the
single most important cause of child poverty in Australia, and has been linked to poorer
developmental outcomes for children (Hand et al. 2011). Jobless families are therefore reliant
on government income supports. In the past couple of decades, many government payments
have become conditional, in an attempt to address concerns about the welfare of children. An
example is compulsory income management or welfare quarantining, which aims to ensure
household expenditure on priority items that meet children’s needs rather than gambling,
pornography, alcohol and junk food, particularly in circumstances where authorities have
concerns about child neglect (Taylor, Stanton & Gray 2012). Such conditionality is directly or
indirectly aimed at shaping parental behaviours and the family environments in which
children grow up.
Community efforts to address family safety and wellbeing
Although not a national policy, a key development addressing children’s safety and wellbeing
through systemic, community-wide interventions to support families was the Northern
Territory Emergency Response (NTER). Initially the NTER aimed to reduce child sexual
abuse, but the high incidence of neglect and the presence of risk factors for other
maltreatment types quickly expanded its focus.4 There is no nationally representative
prevalence study of family dysfunction or child abuse and neglect anywhere in Australia to
allow for comparisons between jurisdictions, nor is there adequate historical data to allow an
assessment of whether levels of child abuse or neglect in the Northern Territory have changed
since the implementation of the NTER. However, the limited qualitative and quantitative data
available suggest that NTER measures have increased awareness of the problem of child
safety within Indigenous communities and provided additional resources to allow for
increased reporting and investigation. At the least, the NTER has not made the situation worse
for families and children; many indicators are pointing in the right direction (Scott & Higgins
2011).
2.2
Public health approach to child protection: implications for research
Improving the systems of protection for children requires a focus on the parental behaviours
that can impair capacity to provide positive parenting and ensure children are safe and
protected from harm (Higgins & Katz 2008). In extreme cases, where parents are affected by
drugs and alcohol, have untreated mental illness or are experiencing inter-parental violence,
their capacity to be attuned or responsive to children’s needs is likely to be diminished.
However, a commitment to a public health approach to child protection requires research to
focus more generally on all families rather than solely on families where children are most at
risk of maltreatment (Scott 2006). It emphasises promoting safe and supportive environments
for all children rather than concentrating on those environments where children may be at risk
of abuse or neglect. The move towards a public health approach to child protection reflects, in
some part, a move in research away from viewing parents who maltreat children as a distinct
psychological category towards viewing them as being at one end of a continuum that
includes all parents (Azar 2002; Belsky 1984; Holden 2010). Children experience varying
levels of risks across this continuum, from warm, positive parenting through to cold,
6
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS
unresponsive, highly neglectful or abusive parenting and exposure to violence occurring in
the family.
A public health perspective on child protection asks us to recognise that levels of parental
warmth and hostile or angry parenting vary across families, and that parent–parent conflict
arises in a broad range of families throughout society. These issues are not distinct as
problems with parenting, and parent–parent conflict may co-occur in risky families (Repetti,
Taylor & Seeman 2002). Therefore, these issues may arise jointly, to varying degrees, across
families in society more generally. Understanding more about the range of family
environments that children experience could lead to important insights that may be useful in
efforts to further promote safe and supportive environments for all children.
In this section, we outline two of the core elements of a safe and supportive family
environment highlighted in previous theoretical and empirical literature. The first relates to
parenting and the second to parent–parent conflict. Extreme problems with parenting and
parent–parent conflict correspond directly to child protection problems. We then discuss
research that links these two elements, describe the development of broad types of families
based on these elements, and consider findings that link these types with child outcomes.
Parenting
Parents embody the family environment for children. Ideally, they provide a secure base from
which children can safely explore the world, and continually provide opportunities for
children to learn about themselves, others and the wider world around them (Holden 2010).
Central to this are the parenting practices that children experience in their day-to-day
interactions with their parents. A safe and supportive environment is one in which children
experience warm, positive interactions with their parents (Petit, Bates & Dodge 1997). In
addition, children thrive in family environments that have well-defined, but not rigid,
boundaries in which parental discipline is consistently applied (Baumrind & Black 1967). By
contrast, interactions characterised by parental anger or hostility towards children point to a
more risky family environment (Repetti et al. 2002).
Research has consistently shown that parenting practices influence child outcomes (Lucas,
Nicholson & Maguire 2011; O’Connor & Scott 2007). Petit and Bates (1989) observed
children at the ages of 6, 12, 24, 36 and 48 months (n = 29) and found a significant positive
relationship between warm parenting and lower child aggression or social withdrawal. Studies
have also shown that parenting in the early years influences adolescent risky behaviour. For
example, Alati et al. (2010) used data from the Mater Hospital Study of Pregnancy (n =
4,158) to show that 14-year-olds were at increased risk of consuming alcohol if their mother
had exerted relatively low control in their parenting when their children were aged 5.
Finally, as children grow it is important that they engage in shared activities with their parents
(Wise 2003). These are important opportunities to develop both cognitive and non-cognitive
skills. For example, shared parent–child engagement in reading (Senechal & LeFevre 2002)
and play (Tamis-LeMonda, Užgiris & Bornstein 2002) has a positive influence on children’s
cognitive, social and emotional development.
Parent–parent conflict
Another important component of the family environment relates to conflict between parents
(Zubrick et al. 2008). Conflict is an inherent part of any normal relationship. Heightened,
recurring conflict is a feature of highly risky family environments and can lead to adverse
psychological and behavioural outcomes for children (Cummings & Davies 2010; Repetti et
al. 2002). Such harmful conflict may include non-verbal and verbal aggression, hostility, and
7
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS
violence against a person or objects. Parental conflict tends to be more problematic for
children if it remains unresolved (Cummings et al. 1991). On the other hand, conflict may
actually be constructive if children witness a meaningful resolution along with clear
explanations about how this was reached (Cummings & Davies 2002; Hetherington 2006).
Moreover, research shows that negative conflict tactics, such as hostility, elicit negative
emotional responses from children, whereas positive conflict tactics, such as calm discussion,
elicit positive emotional responses (Cummings, Goeke-Morey & Papp 2003).
The manner in which parent–parent conflict affects children is the subject of some debate
(Golombok 2000). Conflict can affect children directly by causing them distress. Children
could themselves be drawn into, or become the subject of, arguments and conflict. Conflict
can affect children indirectly through its negative impacts on parenting, and it can provide a
poor model of interpersonal relationships (Amato 2006).
With respect to direct effects, early research highlighted that preschool children exhibit
negative emotional responses following exposure to anger and conflict, both in the home
(Cummings, Zahn-Waxler & Radke-Yarrow 1981) and in laboratory settings (Cummings
1987). More recently, researchers have drawn links between children’s psychological and
physiological responses to parent–parent conflict. Davies et al. (2008) showed that elevated
distress in response to (simulated) parent–parent conflict was associated with higher cortisol
levels, and this association was particularly strong when the child was involved in the
conflict. This is important because physiological processes may provide critical links to the
longer term impacts of emotionally significant events (Davies et al. 2008). With respect to
indirect effects via parenting, a study of 283 families with children aged 8–16 showed
significant indirect relationships between parental conflict and child externalising problems
(where the focus is directed outward to others, i.e., aggression; cf. internalising behaviours
directed toward the self, such as depression, substance misuse, etc.) through lower levels of
parental warmth and lower parental psychological autonomy (higher guilt/anxiety induced
control) (Schoppe-Sullivan, Schermerhorn & Cummings 2007). Schoppe-Sullivan et al. also
found significant indirect relationships between parental conflict and children internalising
problems through lower levels of parental psychological autonomy and behavioural control.
The family as a system: theoretical and empirical research
An approach that jointly considers parenting, parent–child interactions and parent–parent
conflict, rather than separately examining specific elements such as parent–child or parent–
parent relationships, touches closely on the functioning of the family as a system. Family
systems theorists view the family as a distinct unit that can adapt to changing circumstances
to sustain itself and to set and achieve its goals (Broderick 1993). The family system
comprises a number of subsystems, including relationships between parents, relationships
between parents and children, and relationships between children. In healthy families, the
boundary between these subsystems must be well defined but flexible enough to allow
members in one subsystem (for example, children) to draw on resources from another system
(for example, parents) and to allow the family to adapt to change either within or outside the
family system (Cox & Paley 1997). A family systems perspective views problems in families
as often coinciding with problems relating to boundaries between family subsystems
(Minuchin 1978).
Minuchin (1978) argued that the boundary between parents and children ranged along a
spectrum from extremely rigid through to extremely diffuse. Families in which this boundary
is extremely rigid exhibit higher levels of distance between family members; Minuchin
referred to such families as disengaged. From the child’s perspective, these families are
unresponsive, offering little in terms of warmth and support, and generally provide less
8
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS
protection. At the other end of the spectrum are families in which this boundary is extremely
diffuse, which Minuchin referred to as enmeshed. In these families, the relational distance
between parents and children is extremely close; this can lessen autonomy (on the part of
either the parent or the child) and potentially heighten the intensity of parents’ responses to
relatively minor variations in children’s behaviour. Families not at these extremes were
implicitly defined in opposition to both disengaged and enmeshed families, and referred to as
cohesive. Cohesive families exhibit warmth and support for children, maintaining clear
boundaries to ensure that problems do not spill over between subsystems.
At the extreme, there is some degree of overlap between these types of families and child
abuse and neglect. Neglectful families are clearly disengaged, being generally unresponsive to
children’s needs (Gaudin et al. 1996). On the other hand, violence or abuse in families
provides a strong indication that the clear and proper boundary between parents and children
has disintegrated, aligning closely with the idea of enmeshment. In addition, as this boundary
becomes more diffuse (as in more enmeshed families), parental conflict tends to directly
influence the child. Minuchin (1978) referred to instances where parents can deflect conflict
onto children (which he called detouring) or a parent may form a coalition with a child,
thereby involving them in the conflict. This highlights the influence of diffuse boundaries in
facilitating the spillover effect of parental conflict, discussed above.
However, Minuchin (1978) emphasised that these characteristics arise to some extent in all
families. Therefore, his ideas provide us with a framework for considering family issues that
arise at varying levels across the population but are especially problematic at the extremes. A
number of applied research papers have explored Minuchin’s ideas in normative samples, as
outlined in the following section.
Previous research on family environments in normative samples
Kerrig (1995) showed that all three family groups—enmeshed, disengaged and cohesive—
emerged in a normative sample of families with children 6–10 years old (n = 75) and that the
majority of families were cohesive. Kerrig showed that cohesive families had lower levels of
parental conflict and that children in these families perceived conflict as less intense than
children in enmeshed or disengaged families. The study found significant associations
between these family groups and children’s behavioural problems. In particular, children in
disengaged families had the highest externalising problems score and those in enmeshed
(detouring) families had high levels of internalising problems.
Johnson (2003) observed the family functioning of 57 families when children were aged 5–6
and again when children were aged 9–10 (n = 57). Johnson extracted cohesive and enmeshed
family groups at both time points, and a (separate) disengaged family group when children
were in fourth grade (aged 9–10). There was little consistency, however, in group
membership at the two time points. There was no association between family groups
measured when children were aged 5–6 and their externalising problems (aggression and
hyperactivity), either concurrently or prospectively, when children were aged 10–11.
However, children aged 10–11 in enmeshed families (measured concurrently) had more
externalising problems than children in cohesive or disengaged families. In addition, Johnson
found a significant increase over time in externalising problems for children who were always
in cohesive families, and a significant decrease for children who were in a cohesive family
when aged 10–11 but not earlier.
Sturge-Apple, Davies and Cummings (2010) identified cohesive, disengaged and enmeshed
family groups using data from 234 families with children in kindergarten (average age 5
years). In line with previous work, they showed that there were significantly higher levels of
9
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS
parental conflict in enmeshed families compared with either cohesive or disengaged families.
They also highlighted the importance of different dimensions of parenting. For example,
warm, emotionally available parenting was highest in cohesive families and lowest in
disengaged families. They found that externalising problems increased at a significantly faster
rate for children in disengaged families, and that internalising problems increased for children
in both disengaged and enmeshed families. They also found that children in disengaged
families exhibited increased levels of difficulty with classroom engagement (factors such as
attention span and children’s participation in the classroom). Moreover, children in both
disengaged and enmeshed families exhibited increased emotional difficulties associated with
school. This points to potential links between the family environment and cognitive outcomes
as well as social and emotional outcomes.
Research focus
The present study seeks to build on previous research in a number of important ways. First, it
uses the Longitudinal Study of Australian Children (LSAC), which contains a large,
nationally representative sample of children and their families. Previous studies have been
limited in terms of sample size, and this study overcomes this limitation. With this nationally
representative data, we can explore the prevalence of different types of family environments
across society. As noted above, disengaged and enmeshed families exhibit patterns that, at the
extremes, reflect those evident in families where concerns about child protection arise. In
contrast, cohesive families provide examples of a safe and supportive family environment for
children. Understanding more about the prevalence of different family groups using a large
representative sample, and about the relationship between these modes of family functioning
and child outcomes, can inform policy concerning a public health approach to child
protection.
Second, this study breaks new ground in not being restricted to families with two resident
parents. Previous studies have focused on families with two resident parents and excluded
single-parent and separated families, largely because it seeks to explore the interplay of
conflict among parents who live together, parent–child interaction and child outcomes.
However, conflict between parents who do not co-reside may also shape different types of
family environments and influence child outcomes.
This study has three broad, interlinked objectives. First, and central to this report, using a
range of measures of parenting, parent–child interactions and parental conflict, we examine
whether, using a nationally representative dataset of Australian families, we can identify the
different family environments suggested by theory and previous small-scale empirical work.
Second, this study presents an overview of a broad range of indicators of children’s health,
cognitive development, and social and emotional wellbeing. The National Framework for
Protecting Australia’s Children (COAG 2009a) highlights the importance of monitoring a
number of such key indicators, and this report explores a range of measures available in
LSAC that may support this. For example, we are able to focus on outcomes such as
indicators of physical health (for example, underweight), social and emotional wellbeing, and
literacy and numeracy.
The third part of the report draws the previous two parts together in examining the influence
of different family environments on select child outcomes. The broad aim here is to
understand more about which particular aspects of a safe and supportive family, at different
points in time, may influence various child outcomes. Moreover, we consider change over
time both in the family environment and in child outcomes.
The key research questions addressed in this report are:
10
SAFE AND SUPPORTIVE FAMILY ENVIRONMENTS
1. In families with two resident parents and those with a parent living elsewhere, what is
the prevalence of different types of family ‘groups’ or environments (cohesive,
disengaged, enmeshed)?
2. What are the profiles of different family environments in terms of parenting, parent–
child interactions, and parental conflict?
3. What are the social, demographic and economic characteristics of families in these
environments?
4. What are the links between different family environments and child outcomes?
5. To what extent do changes in family environment influence changes in child
outcomes?
11
CLASSIFYING FAMILY ENVIRONMENTS: METHODOLOGY
3 Classifying family environments: methodology
This chapter, and the one following, addresses the first key objective of this report, which is
to identify safe and supportive family environments as well as others that are more risky.
Section 3.1 provides a brief overview of the methodology used to identify different family
environments, Section 3.2 describes the data and sample used in this analysis and Section 3.3
describes the measures used in the study.
3.1
Latent class cluster analysis
Theory posits that families fall into well-defined groups that are qualitatively distinct based
on the nature of the boundary between subsystems of the broader family system (see
Chapter 2). We cannot identify these groups directly but must rely on observed measures
correlated with the unobserved (latent) categorical variable. An appropriate technique for
identifying different family environments is latent class cluster analysis, also known as latent
profile analysis (Vermunt & Magison 2002). Latent class cluster analysis is the analogue of
latent class analysis, when the observed indicators of the latent categorical measure are
continuous. Previous work in this area uses this approach (for example, Sturge-Apple, Davies
& Cummings 2010). One of the main advantages of this approach is that it provides clear
statistical guidance for choosing an optimal number of clusters such that the differences in
observed measures among individuals within clusters is minimised and the difference
between groups is maximised (for more detail see Vermut & Magison 2002). The terms
‘cluster’ and ‘class’ can be used interchangeably in this context and we use ‘class’ in this
report, in line with previous work in this area (for example, Sturge-Apple, Davis &
Cummings 2010). We used the statistical package Mplus (version 3.01) to estimate the latent
class cluster models.
3.2
Data and sample
We use data from the Longitudinal Study of Australian Children (LSAC), conducted in
partnership between the Department of Social Services (formerly the Department of Families,
Housing, Community Services and Indigenous Affairs), the Australian Institute of Family
Studies and the Australian Bureau of Statistics. This is Australia’s first nationally
representative longitudinal study of children. It aims to understand the role of the social,
economic and cultural environment in social, emotional and cognitive development of
children in Australia.
LSAC follows two cohorts of children and their families from all states and territories of
Australia. The baby cohort (B cohort) comprises children who were born between March
2003 and February 2004 (5,107 children in Wave 1), and the kindergarten cohort (K cohort)
comprises children who were born between March 1999 and February 2000 (4,983 children in
Wave 1). Data collection commenced in 2004 and continues every two years. By 2011, four
waves of data were available. A detailed description of LSAC’s recruitment procedure and
study design can be found in Soloff, Lawrence & Johnstone (2005) and Gray & Smart (2009),
respectively.
With these longitudinal data, we have the opportunity to explore the family environment at
different stages of children’s development, from early to middle childhood. In this study, we
use LSAC data collected when study children were aged 2–3 (B cohort, Wave 2), 4–5
(K cohort, Wave 1), 6–7 (B cohort, Wave 4) and 10–11 years (K cohort, Wave 4). Therefore,
we used two waves of data each from the B cohort and the K cohort (see Figure 1). Data were
collected from two resident parents or carers—the primary resident parent (P1) and secondary
resident parent (P2).
12
CLASSIFYING FAMILY ENVIRONMENTS: METHODOLOGY
This selection allows us not only to compare children within the same cohort over an
extended period but also to exploit the cross-sequential nature of the LSAC design to fill in
the gap between these within-cohort comparisons.
Figure 1: Wave selection for families with two resident parents
In each cohort, the first wave of LSAC did not collect data from parents living elsewhere
(PLEs) from the parent who was the primary respondent about the child/family. Therefore, to
ensure the same age spread of study children with a PLE, we also used data from Waves 2
and 3 of the B cohort (children aged 2–3 and 4–5 years) and Waves 2 and 4 of the K cohort
(children aged 6–7 and 10–11 years).
Figure 2: Wave selection for families with a parent living elsewhere (PLE)
In families with a PLE, data from the PLE may be missing because at Wave 1 the parent who
was the primary respondent (primary parent) did not want the PLE to be contacted, or because
at Wave 1 or any subsequent wave the PLE did not wish to respond. Therefore, in our
multivariate analyses, any families with a PLE where all data from the PLE are missing have
been excluded.
In order to consider transitions in the family environment, we selected families in each cohort
who responded at both waves. In addition, we selected only families where the primary carer
was either:

in a couple at both waves or

separated from the study child’s other parent at both waves.
The reason for this is that the measures used in collecting data from these different types of
families are not identical (for example, angry parenting was not collected consistently from
the PLE, and the measure of parental conflict differs for a PLE). This would influence the
consistency of the measurement of the family environment. For example, if a couple
separated between waves, the measures available for that family would alter and the
measurement of the family environment at different waves would not be comparable. Future
work could consider the impact of these changes in family formation, but it is beyond the
scope of this report.
The LSAC sample size included in the current analysis totalled 7,413 families. Families in
which the study child lived with two resident parents at both time points numbered 3,439 in
the B cohort and 3,309 in the K cohort. Families in which there was a single primary
parent/carer responding to the survey and a PLE at both time points numbered 301 in the
B cohort and 364 in the K cohort. (Children living with a sole resident parent and who had a
deceased parent are a very small subgroup—only 11 in the B cohort and 21 in the K cohort—
and were excluded from the analysis.) This was also the base sample that we used in the
analysis in Chapter 5 (however, for some analyses, missing data on some independent
13
CLASSIFYING FAMILY ENVIRONMENTS: METHODOLOGY
variables reduced the sample size further). We provide details of the analysis sample with the
results.
3.3
Measures
In this section, we describe the measures used in the latent class cluster analysis to identify
different family environments. Previous work has adopted a range of approaches to measuring
the family environment. Some researchers attempt to measure dimensions of the family
environment directly either through direct observation of families (for example, Johnson
2003; Sturge-Apple, Davis & Cummings 2010) or using instruments designed to directly
measure aspects of family interactions (for example, Kerrig 1995). A third approach, most
similar to that adopted in this report, uses a range of indicators of parenting, parent–child
interactions and parental conflict to measure a latent categorical variable capturing different
family environments (for example, Davies, Cummings & Winter 2004).
LSAC contains numerous measures of different aspects of parenting and more limited
measures of parental conflict. We chose measures that were most consistent with the
theoretical empirical research. In addition, we chose measures collected consistently across
both time points. Lastly, we chose measures that were, as far as possible, consistently
collected from the primary and secondary resident parents/carers and, where applicable, a
PLE.
The measures we chose were warm parenting and angry parenting. In addition, we included a
measure of parent–child shared activities to capture positive parent–child interactions and
reflect, in part, the extent to which parents are a resource that their children can access.
Finally, we include a measure of parental conflict.
We used confirmatory factor analysis to create standardised composite scores for each of
these constructs. This approach is sensitive to the relative weight that a particular item may
have in relation to the underlying construct, and deals explicitly with measurement error. In
addition, it provides a number of indicators of the validity and reliability of the measure of the
underlying construct. We use the chi-square and root mean square error of approximation
(RMSEA) statistics to measure construct validity, and coefficient H to measure construct
reliability (Holmes-Smith 2012). Chi-square p-values above .05 indicate good construct
validity. This statistic is sensitive to deviations from the normal distribution, and the RMSEA
is a less sensitive alternative measure of construct validity (values < .08 indicate reasonable
construct validity and values < .05 indicate good construct validity).
The primary and secondary resident parents/carers (P1 and P2) and the parents living
elsewhere (PLEs) answered a number of questions relating to warm parenting (for example,
‘How often do you hug or hold this child?’ ‘How often do you tell this child how happy
he/she makes you?’). The composite scores had good validity and reliability. The chi-square
p-value was greater than .05 across all models with the exceptions of P2: B cohort, Wave 2;
and P1 and P2: K cohort, Wave 1. The RMSEA was less than .05 in these cases. The
coefficient H exceeded .76 across all models.
The primary and secondary resident parents/carers (but not the PLEs) answered a number of
questions relating to angry parenting (for example, ‘How often are you angry when you
punish this child?’ ‘How often have you lost your temper with this child?’). The composite
scores had good validity and reliability. The chi-square p-value was greater than .05 across all
models with the exceptions of P1 and P2: B cohort, Wave 4 and K cohort, Wave 4; and P2:
K cohort, Wave 1. The RMSEA was less than .05 in these cases (less than .08 for P2:
K cohort, Wave 4). These constructs had lower reliability (.6–.8), which may be attributed to
14
CLASSIFYING FAMILY ENVIRONMENTS: METHODOLOGY
difficulties in the wording of questions to identify this aspect of parenting while still
maintaining responses to this longitudinal study.
The primary parents and the PLEs stated how often during the week prior to the interview
they had read or told a story to the study child, played indoors or outdoors with the study
child, engaged in music or other creative activities with the study child, or included the child
in everyday activities. The RMSEA was less than .05 in all models. Reliability was higher for
younger study children (.7) but decreased to .5–.6 for older children. As children grow older,
they are less likely to be interacting with their parents in these ways, which probably explains
the lower reliability of this construct. We persisted with using this measure, however, as the
results were consistent with theoretical expectations, although (as will be shown) the sizes of
the effects were lowest of all the indicators.
In families with two resident parents, both parents answered questions relating to parental
conflict (for example, ‘How often is there anger or hostility between your partner and you?’
‘How often do you have arguments with your partner that end up with people pushing, hitting,
kicking or shoving?’). In families with a PLE, the primary resident parent also answered
questions relating to conflict with the PLE. In two-parent families, the chi-square p-value is
greater than .05 across most models except for P2: B cohort, Wave 2. The RMSEA is less
than .05 in all models, and construct reliability is good at around .8.
In the models of parental conflict in families with a PLE, the chi-square p-value is greater
than .05 in the models for the B cohort, Waves 2 and 3, but not in K cohort, Waves 2 and 4.
The RMSEA is also greater than .08 in models for K cohort, Waves 2 and 4. Other indicators
of construct validity offer more promising results (CFI = .98; SRMR < .05)5 but, generally,
these models appear relatively weaker. This could be a function of the nature of conflict
between parents in families with a PLE as children grow older. The measure of conflict
obtained from the PLE is based on responses to a single question about how well the PLE gets
along with the study child’s primary responding parent. Responses range from 1 (very well)
to 5 (very poorly/badly). This is a limited measure and, when coupled with the relatively
poorly fitting models for parental conflict obtained from the primary resident parent, shows
that the measures of parental conflict in these families are not as strong as those in families
with two resident parents. Notwithstanding this, we have sought to maximise the potential of
the data with respect to the research questions considered in this report. Any findings,
however, should be judged within the limitations of the measures we use to identify the latent
categorical measure of family environments.
15
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
4 Classifying family environments: results
In Chapter 2 we described different family environments (characterised by family groups
exhibiting cohesion, disengagement and enmeshment). In this chapter we consider the extent
to which these different family environments can be identified in Australia using a large-scale
nationally representative dataset. This chapter presents the results of a latent class cluster
analysis, described in Chapter 3, and has three components. Section 4.1 presents basic
descriptive information about the number of families in each class. (The Appendix contains
technical details concerning the selection of the classes). Section 4.2 provides detailed
descriptions of the different classes extracted from the latent class cluster analysis with
respect to the factors used to elucidate the classes. This provides insights into the extent to
which the different classes resemble family environments suggested by theory and previous
research. Section 4.3 explores some of the characteristics of the different family environments
in terms of the characteristics of the parents and the broader social environment.
Box 1: Chapter 4 key findings
This chapter presents the results of a latent class cluster analysis seeking to identify different types of family
environments.
Reflecting theoretical prediction and previous research, we identified three broad family environments:
•
cohesive: the largest group of families exhibited average or above-average levels of parental warmth
and parent–child shared activities, and below-average levels of hostile parenting and parental relationship
conflict. These families most resemble those noted for boundaries that are clear but flexible.
•
disengaged: a smaller group of families exhibited below-average levels of parental warmth and parent–
child shared activities, average or below-average levels of parental conflict and above-average levels of
hostile parenting
•
enmeshed: a relatively small group exhibited average levels of parental warmth, and higher than
average levels of conflict in the relationship between parents.
We identified these family environments across a broad age range of study children, both in families with
two resident parents and in families with a parent living elsewhere from the primary carer. Having two or
more siblings was significantly associated with a higher likelihood of being in a relatively more disengaged
family rather than in a relatively more cohesive family.
4.1
Identifying different family environments
This section addresses Research Question 1 (‘In families with two resident parents and those
with a parent living elsewhere (PLE), what is the prevalence of different types of family
environment?’) and presents some descriptive information about the prevalence of different
family environments. The latent class analysis showed that three distinct classes were
observed in the data (see Appendix for technical details). Table 2 shows the number and
percentage of cases in each of the three classes extracted from the latent class cluster analysis
in families with two resident parents. In families with a study child aged 2–3 years, there is
one large group of families (n = 2,433; 71 per cent of all families) and two smaller groups of
relatively similar size (n = 576; 17 per cent and n = 430; 12 per cent). Similar patterns arise in
families with a study child aged 6–7 or 10–11 years. In families with a study child aged 4–5
years there is also a large group, but this is smaller than for other families (n = 2,068; 63 per
cent), and the next largest group is substantially larger than that in families with study
children of other ages (n = 761; 23 per cent).
16
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Table 2:
2–3 years
4–5 years
6–7 years
10–11 years
Final class counts and proportions: families with two resident parents
N
%
N
%
N
%
N
%
Class 1
Class 2
Class 3
Total
2,433
576
430
3,439
70.7
16.8
2,068
480
62.5
14.5
483
2,557
14.0
74.4
434
2,428
13.1
73.4
12.5
761
23.0
399
11.6
447
13.5
100.0
3,309
100.0
3,439
100.0
3,309
100.0
Source: LSAC B cohort, Waves 2 and 4; K cohort, Waves 1 and 4.
Table 3 shows the number and percentage of cases in each of the three classes extracted from
the latent class cluster analysis in families with a PLE. The smallest group is Class 2 for
families with a study child aged 4–5 years (n = 28; 9 per cent). The two-class solution yielded
a small group with 44 cases, which is very similar in magnitude to Class 3 in the three-class
solution. Therefore, the three-class solution yields two groups that are similar (in magnitude)
to the two-class solution, with a third group that may prove useful in subsequent analyses.
The distribution across classes of families with study children aged 6–7 and 10–11 years old
(K cohort) and with a PLE differs somewhat from the distribution of families with younger
study children and in families with two resident parents. Families with a study child aged 6–7
years are split comparatively equally across the three classes. The smallest group is Class 2 (n
= 113; 31 per cent) and the largest is Class 3 (n = 131; 36 per cent). In families with a study
child 10–11 years old, there are two groups of relatively equal magnitude (around 40 per cent
of families), and a third group comprising just less than 20 per cent of families.
Table 3:
Final class counts and proportions: families with a PLE
Class 1
2–3 years
4–5 years
6–7 years
10–11 years
N
%
N
%
N
%
N
%
Class 2
70
231
23.3
76.7
232
28
77.1
9.3
120
113
33.0
31.0
154
146
42.3
40.1
Class 3
Total
–
301
–
100.0
41
301
13.6
100.0
131
364
36.0
100.0
64
364
17.6
100.0
Source: LSAC B cohort, Waves 2 and 3; K cohort, Waves 2 and 4.
These results raise questions about the normative group in different family structures. Among
families with two resident parents, there is a group containing the large majority of families,
which most likely represents the normative family group. It is clear that this is not the case for
all families with a PLE considered here, where the largest group changes from 2–3 (Class 2)
to 4–5 years (Class 1), and then disperses with three even groups at age 6–7, and two equally
large groups at age 10–11 (Class 1 & 2). Note, however, that the sample size for families with
a PLE is substantially smaller, and this may influence the results. It is also important to note
that the patterns in families with a PLE are not consistent across families with study children
of different ages. Understanding more about the nature of these different classes will provide
further insights, and we turn to this in the following section.
17
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
4.2
Latent class profiles
This section addresses Research Question 2 (‘What are the profiles of different family
environments in terms of parenting, parent–child interactions, and parental conflict?’) and
presents a descriptive overview of the groups of families extracted from the latent class
cluster analysis. Theoretical and empirical work has established three broad groups of
families. A first group, which is high on cohesion, tends to be characterised by warm
supportive relationships with clear but flexible boundaries, which shelter children from
potential problems in the parental system while maintaining access to supportive resources. A
second group of families, high on disengagement, has relatively more rigid boundaries and
thus tends to offer lower levels of parental warmth and support to children. A third group of
family environments, high on enmeshment, tends to have more diffuse boundaries between
the parent–parent and the parent–child relationships. This group is characterised by higher
levels of parental conflict, to which children tend to be exposed and in which they are likely
to become embroiled.
In our analysis, we highlight those groups of families that tend to have levels of particular
factors closely resembling the profiles of one of these theoretical constructs. It is important to
understand that there are no objective thresholds here. Rather, differences are understood in
relation to the average level of any particular factor, which in the broader population tends to
be high for positive factors and low for more problematic factors. What this means is that the
vast majority of families sit around the middle of this spectrum, with some families tending
toward either end of the spectrum but not approaching, or even coming close to, the extreme
ends.
It is important to note that these different groupings are theoretical devices that we can use to
organise the tremendously complex sets of interrelationships and dynamics across families in
the broader population. For this reason, these characteristics will not manifest perfectly in any
single family. Rather, families will tend to resemble, to a greater or lesser extent, one
particular type of family environment more than another.. Therefore, these groupings function
entirely as a tool to identify similar sets of behaviours and dynamics within particular groups
of families. There are clear limitations here. In particular, it is important to note that no family
environment can be comprehensively understood using a finite set of factors. It is also
important to note that—as we will show later—these are not static groups; families can and
do change over time.
The previous section showed that the latent class cluster analysis extracted four classes for
families with two resident parents and relatively young study children (2–3 and 4–5 years
old), and three classes for families with relatively older study children (6–7 and 10–11 years
old). In families where there is a PLE, the latent class cluster analysis extracted two classes
for families in which the study child was aged 2–3 years, and three classes for all other ages.
The rest of this section describes the identified groups of families with two resident parents in
terms of the factors used in the latent class cluster analysis (warm parenting, angry parenting,
parental conflict and parent–child shared activities), and ascribe labels on the basis of these
descriptive analyses. We then repeat this for families with a PLE (omitting angry parenting, as
LSAC does not consistently collect this data from the PLE). Following this, in Section 4.3, we
explore the characteristics of these groups using multinomial logistic regression for both
family types separately.
Families with two resident parents
In this section we describe the profile of families in each of the three groups we extracted
from the latent class cluster analysis with respect to the factors used in the models (warm
18
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
parenting, angry parenting, parent–parent conflict and parent–child shared activities). In our
presentation of the results, we highlight the idea that these groups of families are positioned
along a theoretically hypothesised range of boundaries between parents and children. Relative
positions on this range are associated with varying levels of factors related to parenting and
parent–parent relationships. In the centre are families with levels that tend toward the average,
whereas families toward either end of the boundary range have relatively higher or lower
levels of particular factors, as described in the previous section (see also Section 2.2).
Figure 3 shows the average standardised factor scores for parental warmth, angry parenting,
parental conflict and parent–child shared activities in the four groups extracted from the latent
class cluster analysis for families with study children aged 2–3 years (B cohort, Wave 2).
Note that these scores have been standardised to have an average of zero. Therefore, values
higher than zero are above average, and values lower than zero are below average.
Families in the largest group (Class 1) have average levels of parent–child shared activities
and above-average parental warmth from both the primary and secondary parents. In addition,
there are lower than average levels of angry parenting and parental conflict from both parents.
These families most closely resemble cohesive families located toward the centre of the
boundary range, as described in previous research.
Families in the second group (Class 2) have noticeably higher levels of parental conflict,
suggesting that they tend toward the more enmeshed end of the boundary range. The primary
parent has below-average warmth and above-average angry parenting, but the secondary
parent has average levels of parental warmth and anger. These families display below-average
levels of parent–child shared activities. There is an element of disengagement by the primary
parent, but the most striking feature relates to parental conflict, which is a key defining
feature of families positioned toward the enmeshed end of the boundary range described in
prior theoretical and empirical work.
Families in the third group (Class 3) tend toward the more disengaged end of the boundary
range. They have lower than average levels of parent–child shared activities, lower parental
warmth and high levels of angry parenting.
19
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Figure 3: Average standardised factor scores for parental warmth, angry parenting, parental
conflict and parent–child shared activities in disengaged (Class 3), cohesive (Class 1)
and enmeshed (Class 2) families: B cohort 2–3 years
Notes:
Weights applied; P1=parent 1; P2=parent 2; HH=household (children could engage in shared activities with P1
and/or P2).
Source: LSAC B cohort, Wave 2.
Figure 4 shows the results for families with a study child 4–5 years old. Families in Class 1
have above-average parent–child shared activities and parental warmth, and below-average
parental conflict and angry parenting. These families resemble cohesive families located
toward the centre of the boundary range. Class 3 families have below-average parent–child
shared activities and parental warmth, combined with above-average levels of angry
parenting, and thus tend toward the more disengaged end of the boundary range. Class 2
families have above-average parental conflict and average levels of parental warmth and
parent–child shared activities, and thus tend toward the more enmeshed end of the boundary
range.
20
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Figure 4: Average standardised factor scores for parental warmth, angry parenting, parental
conflict and parent–child shared activities in disengaged (Class 3), cohesive (Class 1)
and enmeshed (Class 2) families: K cohort 4–5 years
Notes:
Weights applied; P1=parent 1; P2=parent 2; HH=household (children could engage in shared activities with P1
and/or P2).
Source: LSAC K cohort, Wave 1.
In families with older study children (6–7 and 10–11 years old), the latent class cluster
analysis extracted three groups, and descriptive analysis reveals that these correspond to
cohesive, disengaged and enmeshed families. Figure 5 shows the average standardised factor
scores for parental warmth, angry parenting, parental conflict and parent–child shared
activities in the three groups extracted from the latent class cluster analysis for families with
study children aged 6–7 years (B cohort, Wave 4). Class 1 families exhibit above-average
parent–child shared activities and parental warmth; although the differences are not
substantially above average, they are significant. In addition, they have below-average levels
of parental conflict and angry parenting. Relative to the other groups of families shown in
Figure 5, these families are located toward the centre of the boundary range. Class 3 families
have below-average parent–child shared activities and parental warmth, and above-average
angry parenting, which are factors closely associated with families lying toward the more
disengaged end of the boundary range. Class 2 families show substantially higher levels of
parental conflict along with average levels of parental warmth and parent–child shared
activities, resembling the profile of families positioned toward the enmeshed end of the
boundary range.
21
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Figure 5: Average standardised factor scores for parental warmth, angry parenting, parental
conflict and parent–child shared activities in disengaged (Class 1), cohesive (Class 2)
and enmeshed (Class 3) families: B cohort 6–7 years
Notes:
Weights applied; P1=parent 1; P2=parent 2; HH=household (children could engage in shared activities with P1
and/or P2).
Source: LSAC B cohort, Wave 4.
The results for families with study children aged 10–11 years, shown in Figure 6, are very
similar to those in Figure 5 for families with study children aged 6–7 years. One key
difference is that levels of angry parenting in Class 1 (which resembles the profile of families
at the disengaged end of the boundary range) are higher for both parents. Profiles associated
with families in Class 2 resemble the profile of families in the middle (cohesive) end of the
boundary range and in Class 3 resemble the profile of those at the enmeshed boundary range.
22
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Figure 6: Average standardised factor scores for parental warmth, angry parenting, parental
conflict and parent–child shared activities in disengaged (Class 1), cohesive (Class 2)
and enmeshed (Class 3) families: K cohort 10–11 years
Notes:
Weights applied; P1=parent 1; P2=parent 2; HH=household (children could engage in shared activities with P1
and/or P2).
Source: LSAC K cohort, Wave 4.
Families with a parent living elsewhere
This section describes the groups extracted from the latent class cluster analysis of families
with a parent living elsewhere (PLE). Figure 7 shows the average standardised factor scores
for parental warmth, parental conflict and parent–child shared activities in the two groups
extracted from the latent class cluster analysis for families with a PLE (B cohort, Wave 2).
Note that the scores have been standardised to have an average of zero. Therefore, values
higher than zero are above average, and values lower than zero are below average.
Families in the largest group (Class 2) have above-average levels of parental warmth from the
primary parents, and slightly below-average warmth from the PLE. Both parents report
slightly above-average levels of parent–child shared activities. In addition, both parents report
average levels of parental conflict. These families clearly resemble the cohesive families,
located toward the centre of the boundary range, described in previous research. In the second
group of families (Class 1) the primary carer (but not the PLE) reports lower average levels of
parental warmth, and both parents report lower average levels of parent–child shared
activities. These families closely resemble families toward the disengaged end of the
boundary range. The analysis did not identify any families tending toward the more enmeshed
end of the boundary range.
The latent class cluster analysis for families with a PLE when children were aged 4–5 years
extracted three groups of families. Figure 8 shows the average standardised factor scores for
parental warmth, parental conflict and parent–child shared activities for each of these three
groups. Families in the largest group (Class 1) most resemble those toward the centre of the
23
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
boundary range. They have average levels of warm parenting and parent–child shared
activities, and below-average levels of parental conflict.
Figure 7: Average standardised factor scores for parental warmth, parental conflict and parent–
child shared activities in disengaged (Class 1) and cohesive (Class 2) families with a
PLE: B cohort 2–3 years
Notes:
Weights applied; P1=parent 1; PLE=parent living elsewhere from the primary carer.
Source: LSAC B cohort, Wave 2.
24
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Figure 8: Average standardised factor scores for parental warmth, parental conflict and parent–
child shared activities in disengaged (Class 2), cohesive (Class 1) and enmeshed
(Class 3) families with a PLE: B cohort 4–5 years
Notes:
Weights applied; P1=parent 1; PLE=parent living elsewhere from the primary carer.
Source: LSAC B cohort, Wave 3.
Another group (Class 3) is characterised by relatively low levels of primary parental warmth
and parent–child shared activities (PLE–child shared activities are not available in Wave 3 of
the B cohort). Families in this group are positioned toward the more disengaged end of the
boundary range, at least with respect to the primary parent. Families in the last group
(Class 2) have above-average levels of parental conflict, along with slightly above-average
levels of parental warmth and parent–child shared activities. These patterns reflect those
associated with families tending toward the enmeshed end of the boundary range.
Figure 9 shows results for families with a PLE when the study child was 6–7 years old. Again,
the latent class cluster analysis extracted three groups of families. Families in the largest
group (Class 3) most closely resemble cohesive families (above-average parental warmth and
below-average parental conflict). In another group (Class 1), the primary parent has belowaverage parental warmth and parent–child shared activities, suggesting that families in this
group are positioned toward the disengaged end of the boundary range. In these families, the
PLE has average levels of parental warmth and parent–child shared activities. Families in the
remaining group (Class 2) exhibit patterns that resemble families at the more enmeshed end of
the boundary range. Both the primary parent and the PLE report above-average levels of
parental conflict, along with above-average primary parental warmth and above-average
PLE–child shared activities.
25
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Figure 9: Average standardised factor scores for parental warmth, parental conflict and parent–
child shared activities in disengaged (Class 1), cohesive (Class 3) and enmeshed
(Class 2) families with a PLE: K cohort 6–7 years
Notes:
Weights applied; P1=parent 1; PLE=parent living elsewhere from the primary carer.
Source: LSAC K cohort, Wave 2.
The final latent class cluster analysis relates to families with a PLE when the study child was
10–11 years old (Figure 10). The patterns are very similar to those displayed in Figure 9. Three
groups emerge from the analysis, closely relating to the centre (cohesive, Class 1), the
disengaged end (Class 2), and the enmeshed end of the boundary range (Class 3). The most
notable features of these groups are: the cohesive families have above-average warmth,
families located toward the disengaged end of the boundary range have below-average
warmth and parent–child shared activities, while families located toward the enmeshed end of
the boundary range have above-average parental conflict and above-average parental warmth.
26
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Figure 10: Average standardised factor scores for parental warmth, parental conflict and parent–
child shared activities in disengaged (Class 2), cohesive (Class 1) and enmeshed
(Class 3) families with a PLE: K cohort 10–11 years
Notes:
Weights applied; P1=parent 1; PLE=parent living elsewhere from the primary carer.
Source: LSAC K cohort, Wave 4.
4.3
Characteristics of families in each latent class
This section addresses Research Question 3 (‘What are the social, demographic and economic
characteristics of families in these environments?’) and presents the results from a
multinomial logistic regression analysis to assess the characteristics of the families in each of
the groups extracted from the latent class cluster analysis. The dependent variable is a
categorical variable indicating group membership, and the reference group is cohesive
families. The other two categories relate to families located either toward the disengaged end
or the enmeshed end of the boundary range. In this analysis, we restrict the covariates to those
relating to parents, the family, and the area in which the family lives. We introduce factors
relating directly to the study child in analyses focusing on child outcomes. We include the
following covariates in our models:

region: 1 = metropolitan [reference]; 2 = rural or regional area

SEIFA disadvantage bottom quintile: 1 = yes; 0 = no [reference]

parental education: 1 = one or both parents have a degree [reference]; 2 = no parent in
family has Year 12 or equivalent; 3 = one or both parents have Year 12 or equivalent
but none has a degree or higher

jobless family: 0 = no [reference]; 1= yes

lowest household income quintile: 0 = no [reference]; 1= yes

mother’s age at first birth: 0 = 25 years and older [reference]; 1 = under 25 years
27
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS

older siblings in family: 0 = no [reference]; 1 = one older sibling; 2 = two or more
older siblings

younger siblings in family: 0 = no [reference]; 1 = one younger sibling; 2 = two or
more younger siblings (except B cohort, Wave 2).
In addition, for families with a PLE we include a categorical variable measuring how often
the study child spends time with the PLE (1 = at least once per week [the reference category];
2 = 4–26 times per year; 3 = no more than twice per year).
Table 4 shows the relative risk ratios (RRRs) from the regression analysis, for families with
two resident parents. RRRs are akin to odds ratios in a logistic regression. An RRR greater
than 1 indicates a positive difference, an RRR less than 1 indicates a negative difference, and
an RRR equal to 1 indicates no difference. The most consistent and pronounced result across
all models relates to the number of older and younger siblings in the family. In general,
children with more siblings are more likely to live in families located toward the disengaged
end of the boundary range.
Looking at the larger group of families positioned toward the disengaged end of the boundary
range, we can see that having two or more younger siblings increases the likelihood of being
in a disengaged family for children aged 4–5, 6–7, and 10–11 years. Younger siblings may be
likely to absorb relatively more of their parents’ attention, which may explain these results.
The effects of older siblings in the disengaged family group are somewhat mixed. For
children aged 4–5 there is no significant association, whereas for children aged 6–7 there is a
significant positive association. Children aged 10–11 with two or more older siblings are
significantly more likely to be in a family located toward the disengaged end of the boundary
range than children with no older siblings, but this is a relatively small group of children (n =
51).
With respect to other factors in the model, disengaged group membership is unrelated to a
number of socioeconomic factors, including the level of disadvantage in the area, parental
education, family joblessness and income. However, relatively low family income is
positively associated with membership of the enmeshed group for children 6–7 and 10–11
years old. In addition, mid-level parental education is positively associated with membership
of the disengaged family group in the models for children 2–3 years old, but this is a singular
finding and is difficult to explain. Children aged 6–7 living in rural or regional areas are less
likely to be in a more enmeshed family, whereas children aged 10–11 living in a rural or
regional area are less likely to be in a more disengaged family. While both these findings are
positive, there is no consistent pattern across waves and between cohorts with respect to
region.
28
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Table 4:
Relative risk ratios from multinomial logistic regression predicting membership of
disengaged or enmeshed relative to cohesive: families with two resident parents
Disengaged/rigid boundaries (compared with cohesive)
2–3 years
4–5 years
6–7 years
10–11 years
Regional or rural area
1.1
1.0
0.8
0.7*
SEIFA disadvantage bottom quintile
0.8
1.1
1.2
1.1
No parent Y12
1.1
0.9
1.0
1.0
No parent degree
1.3*
0.8
1.0
0.8
Jobless household
0.6
1.3
1.4
1.7
Lowest income quintile
1.1
1.0
1.4
1.1
Mother under 25 at first birth
0.9
0.8
1.2
1.0
1 older sibling
1.2
1.2
1.5**
0.9
2 or more older siblings
1.2
1.2
1.6**
0.6*
1 younger sibling
1.2
1.3*
1.9***
1.7**
2 or more younger siblings
–
1.6*
2.0**
2.0***
Enmeshed/diffuse boundaries (compared with cohesive)
4–5 years
6–7 years
10–11 years
Regional or rural area
0.9
1.0
0.6***
0.8
SEIFA disadvantage bottom quintile
0.8
1.0
1.1
0.9
No parent Y12
0.8
1.0
1.1
1.3
No parent degree
0.8
0.9
1.1
1.1
Jobless household
1.0
0.7
0.4
1.0
Lowest income quintile
1.3
1.0
1.8***
1.6*
Mother under 25 at first birth
1.0
0.9
1.0
1.0
1 older sibling
1.1
1.0
0.7*
1.2
2 or more older siblings
1.1
0.9
1.0
1.3
1 younger sibling
0.8*
1.0
0.9
1.2
2 or more younger siblings
0.9
1.0
0.9
1.2
Number of observations
3,200
3,303
3,237
3,027
Notes:
2–3 years
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC B cohort, Waves 2 (2–3 years) and 4 (6–7 years); K cohort, Waves 1 (4–5 years) and 4 (10–11 years).
Table 5 reports the relative risk ratios (RRRs) from multinomial regression analyses for
families with a PLE. For study children 2–3 and 4–5 years old, having two or more older
siblings is positively associated with living in a family positioned toward the disengaged end
of the boundary range. For children aged 4–5, it is also positively associated with living in a
family located toward the enmeshed end of the boundary range. In general, having a large
number of siblings has a similar effect for children in families with a PLE as it does for those
living in families with two resident parents.
29
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
Table 5:
Relative risk ratios from multinomial logistic regression predicting membership of
disengaged or enmeshed relative to cohesive families: families with a PLE
Disengaged/rigid boundaries (compared with cohesive)
2–3 years
4–5 years
6–7 years
10–11 years
Regional or rural area
0.5
0.5
0.6
0.7
SEIFA bottom quintile
0.7
1.4
0.8
1.2
No parent Y12
0.7
0.9
1.7
1.8
Jobless household
0.6
1.2
0.8
0.9
Lowest income quintile
1.7
0.5
0.9
1.0
Mother under 25 at first birth
0.8
0.6
0.7
0.9
1 older sibling
1.6
4.5*
0.7
1.0
2 or more older siblings
3.0*
6.2*
1.6
1.2
1 or more younger siblings
1.3
2.5
1.4
1.2
Child sees PLE 4–26 times per year
1.1
0.2
1.0
1.3
Child sees PLE no more than twice per year
0.5
0.5
0.6
0.7
Enmeshed(a)/diffuse
boundaries (compared with cohesive)
2–3 years
4–5 years
6–7 years
10–11 years
Regional or rural area
–
0.5
0.8
0.7
SEIFA bottom quintile
–
0.7
1.0
1.3
No parent Y12
–
0.8
1.4
1.2
Jobless household
–
1.2
0.9
0.6
Lowest income quintile
–
0.8
0.7
0.6
Mother under 25 at first birth
–
0.8
1.6
1.9
1 older sibling
–
2.1
1.3
1.1
2 or more older siblings
–
4.2*
1.8
2.0
1 or more younger siblings
–
2.0
0.7
1.4
Child sees PLE 4–26 times per year
–
2.6
1.7
1.3
Child sees PLE no more than twice per year
–
1.1
1.3
1.7
Number of observations
228
(a)
Enmeshed class not extracted for families with study child 2–3 years.
Notes:
Weights applied; *** p < .001; ** p < .01; * p < .05.
189
286
260
Source: LSAC B cohort, Waves 2 (2–3 years) and 3 (4–5 years); K cohort, Waves 2 (6–7 years) and 4 (10–11 years).
4.4
Summary
This chapter has presented the results of a latent class cluster analysis seeking to identify
different types of family environments. Reflecting theoretical prediction and previous
research, we identified three broad family groups. The largest group of families had average
or above-average levels of parental warmth and parent–child shared activities, coupled with
below-average parental conflict and angry parenting. These families most resemble those
described in previous work as having boundaries that are clear but flexible, and referred to as
cohesive. We also identified a group of families with below-average parental warmth and
parent–child shared activities. These also had average or below-average levels of parental
conflict but tended to have high levels of angry parenting. These families closely resembled
those hypothesised to be located toward the more disengaged end of the boundary range.
Finally, we identified a group of families noted for high levels of parental conflict together
with average or above-average levels of parental warmth and parent–child shared activities.
These families most closely resemble those identified in previous theoretical and empirical
research as being located toward the enmeshed end of the boundary range.
30
CLASSIFYING FAMILY ENVIRONMENTS: RESULTS
We identified these family environments across a broad age range of study children, and both
in families with two resident parents and in families with a PLE. Having identified these
family environments, we are interested to know how they relate to children’s outcomes. The
next chapter presents a descriptive overview of a broad range of indicators of children’s
health, development and wellbeing.
31
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
5 Family environments and child outcomes
This chapter considers the links between the family groups described in the previous chapter
and the child outcomes described in Chapter 2. This chapter thus addresses Research
Question 4 (‘What are the links between different family environments and child outcomes?’)
and the third key objective of the report, which is to understand more about the influence of
the family environment on child outcomes. We consider the associations between child
outcomes and the family environment, measured contemporaneously. For example, we
compare outcomes for children 2–3 years old (Wave 2, B cohort) across the three family
groups extracted from the analysis of families with study children aged 2–3 (Wave 2,
B cohort), and so on. In some cases this is not possible, and in those cases we have chosen the
most recent prior measure of the family environment.
In this chapter, the statistical comparisons focus on differences between family groups within
specific age brackets. For example, we compare outcomes for children aged 2–3 in families
high on enmeshment with those in families high on cohesion; however, we do not compare
outcomes for children of different ages (for example, aged 2–3 years compared with 4–5
years) either in the same type of family group or across types of family groups. Changes over
time are considered in the following chapter.
Using all available, comparable data we examine the following outcomes:
1.
2.
3.
4.
5.
6.
the proportion of children underweight or overweight/obese
the proportion of children with one or more injuries in the past 12 months
SDQ prosocial score
SDQ total problem score
NAPLAN reading scores
NAPLAN numeracy scores.
Together, these measures encompass indicators of children’s health, cognitive development,
and social and emotional wellbeing.
We first present a descriptive analysis of these outcomes, comparing the average level for
each outcome in different family environments. In addition to these descriptive analyses, we
present the results of multivariate analysis considering the association between child
outcomes and family environment controlling for a number of factors relating to children,
their parents and families, and the wider social setting they live in. The variables included are:

family environment: 1 = cohesive [reference]; 2 = disengaged; 3 = enmeshed

region: 1 = metropolitan [reference]; 2 = rural or regional area

socioeconomic position (SEP) bottom quartile: 0 = no [reference]; 1 = yes

mother aged under 25 years at first birth: 0 = no [reference]; 1 = yes

older siblings in family: 0 = no [reference]; 1 = one older sibling; 2 = two or more
older siblings

younger siblings in family: 0 = no [reference]; 1 = one younger sibling; 2 = two or
more younger siblings

child gender: 1 = boy [reference]; 2 = girl

general good health: continuous scale

child living with disability: 1 = no [reference]; 2 = yes
32
FAMILY ENVIRONMENTS AND CHILD OUTCOMES

child Indigenous status: 1 = no [reference]; 2 = yes

birth weight: continuous scale, standardised.
The six outcomes listed above take different forms, and this influences the choice of
regression analysis technique. Weight status (outcome 1) is a categorical variable with three
values, and we use multinomial logistic regression to analyse this outcome (normal weight is
the base category). The proportion of children with one or more injuries (outcome 2) is a
binary variable, and we use logistic regression for the analysis of this outcome. The remaining
outcomes (3–6) are continuous, and we used Ordinary Least Squares (OLS) regression to
analyse these. All analyses were weighted to control for the clustering and stratification of the
sample, as well as to correct for non-response bias. The results are presented in three sections.
Section 5.1 presents the results for outcomes 1 and 2, relating to children’s health. Section 5.2
presents the results for outcomes 3 to 5, relating to children’s social and emotional wellbeing .
Section 5.3 presents the results for outcomes 6 and 7, relating to children’s cognitive
development.
Box 2: Chapter 5 key findings
This chapter presents the results from bivariate and multivariate analyses considering concurrent
associations between family environment and children’s outcomes.
There were very few significant associations between family environment and children’s health outcomes.
Significant results were restricted to children 2–3 years living in families with two resident parents.

Children aged 2–3 years in families lying toward the enmeshed end of the boundary range were
significantly more likely to be underweight (than normal weight).

Children aged 2–3 years in families located toward the disengaged end of the boundary range were
significantly more likely to have two or more injuries per year.
The family environment (as measured in this report) was most strongly associated with children’s social
and emotional wellbeing.

In families with two resident parents, children in families positioned toward the disengaged end of the
boundary range had significantly lower levels of prosocial behaviour, higher levels of total problem
behaviour, and higher levels of externalising problem behaviour when compared to children from more
cohesive families.

Results were very similar for children in families with a parent living elsewhere (PLE). There were also
significant associations highlighting negative social and emotional outcomes for children in enmeshed
families, but these were not as pronounced compared with the results for more disengaged families.
There were less consistent, and fewer significant, associations between family environment and children’s
cognitive development. In families with two resident parents:
5.1

Children in families located toward the disengaged end of the boundary range had, on average, lower
Year 3 and Year 5 NAPLAN reading scores. However, the Year 3 result was not significant after
controlling for other factors.

Children in families situated toward the disengaged end of the boundary range had, on average, lower
Year 5 NAPLAN numeracy scores, even after controlling for other factors. Patterns were very similar
across family environments for children in families with a PLE.
Children’s health
Weight status
In this section, we present the results of analyses examining the associations between
children’s weight status and family environment measured concurrently. The indicator we
used was body mass index (BMI), defined as weight/height squared (kg/m2). This is a
33
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
common indicator used to identify people who are underweight or overweight, or obese.
Although not a direct measure of body fat, BMI remains one of the most widely used
measures in the study of obesity (Sweeting 2007). Consistent with Wake and Maguire (2012),
we used cut-off points for overweight and obesity derived by Cole et al. (2001) and cut-off
points for underweight derived by Cole et al. (2007). We used data on children’s BMI from
Waves 2–4 in the B cohort and Waves 1–4 in the K cohort. BMI data from children aged 0–1
years were not available, due to the difficulty of measuring the height of very young children.
Figure 11 shows, across the different family groups, the proportion of children aged 2–3 years
to 10–11 years who were underweight or overweight/obese in families with two resident
parents. Figure 12 reports the same for families with a PLE. Broadly, the proportion of
children who were either underweight or overweight/obese was very similar across different
family types, and very similar to the average proportions reported in previous research using
the entire LSAC sample (Wake & Maguire 2012).
The only instance of a difference (with non-overlapping confidence intervals) related to
children aged 2–3 years in families with two resident parents located toward the enmeshed
end of the boundary range. These children were more likely to be underweight than children
in families located toward the middle of the boundary range (cohesive families).
There were no notable differences across different family groups (within each child age band)
in the proportion of children in families with a PLE who were underweight or
overweight/obese. Note that a very small proportion of children aged 4–5 years in enmeshed
families were underweight, and that no children aged 10–11 years in this family type were
underweight.
Table 6 reports the results for children in families with two resident parents from a series of
bivariate and multivariate multinomial regression analyses considering associations between
children’s weight status and the family environment with and without other control variables.
Table 7 reports results for children in families with a PLE. In both cases, we estimated models
for children 2–3, 4–5, 6–7 and 10–11 years of age. The bivariate analyses confirm the
description provided in Figure 11 showing a significant difference in the likelihood of children
aged 2–3 years being underweight in cohesive families compared with children aged 2–3
years in enmeshed families, even when controlling for other factors (p < .05). The bivariate
and multivariate analyses reveal no significant differences across family environments in
families with a PLE.
34
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Figure 11: Average percentage of children underweight or overweight/obese in different family
groups: children 2–3 years, 4–5 years, 6–7 years and 10–11 years: families with two
resident parents
Source: LSAC B cohort, Waves 2 and 4; K cohort, Waves 1 and 4.
Figure 12: Average percentage of children underweight or overweight/obese in different family
groups: children 2–3 years, 4–5 years, 6–7 years and 10–11 years: families with a PLE
Source: LSAC B cohort, Waves 2 and 3; K cohort, Waves 2 and 4.
35
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Table 6:
Relative risk ratios from multinomial logistic regression predicting children’s weight problems: families with two resident parents
2–3 years
Overweight/
Underweight
obese
4–5 years
Overweight/
Underweight
obese
6–7 years
Overweight/
Underweight
obese
10–11 years
Overweight/
Underweight
obese
Bivariate results
Disengaged (more rigid boundaries)
1.1
Enmeshed (more diffuse boundaries)
1.7**
Number of observations
1.1
1.1
1.0
1.1
3,381
1.1
1.7
1.0
1.4
3,291
1.2
1.0
1.0
1.2
3,394
0.8
1.0
3,191
Multivariate results
Disengaged (more rigid boundaries)
1.0
1.0
1.1
1.1
1.6
1.2
0.9
0.7*
Enmeshed (more diffuse boundaries)
1.6*
1.0
1.0
1.1
1.3
1.0
1.1
0.9
Regional or rural area
0.9
1.0
1.0
1.0
0.9
0.8
1.1
1.0
SEP: bottom quartile
1.1
1.4**
0.7
1.6***
0.9
1.3*
0.9
1.7***
Mother under 25 at first birth
1.3
1.1
1.3
1.1
1.2
1.1
1.0
1.4**
1 older sibling
1.3
0.9
1.4
0.7*
1.1
0.8
1.0
0.8*
2 or more older siblings
1.4
1.2
1.2
0.7*
1.0
1.1
1.3
0.9
1 younger sibling
1.1
0.9
1.4
1.0
0.9
0.8*
1.3
0.7**
2 or more younger siblings
–
–
0.5
0.9
1.0
0.6*
1.0
0.7*
Girl
1.0
1.1
1.3
1.2*
0.9
1.1
1.7**
1.0
Generally healthy
0.8***
1.1
0.7**
1.1
0.7**
0.8*
0.6***
0.6***
With disability
1.4
1.4
–
–
2.0*
1.5
1.3
1.5*
Indigenous
0.7***
1.3***
0.6***
1.3***
0.7***
1.2***
0.8***
1.1
Birth weight
0.3
0.6
0.6
1.6
1.9
1.2
2.6
1.1
Number of observations
Notes:
3,364
3,249
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC B cohort, Waves 2 and 4; K cohort, Waves 1 and 4.
36
3,316
3,141
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Table 7:
Relative risk ratios from multinomial logistic regression predicting children’s weight problems: families with a PLE
2–3 years
Underweight
Overweight/obese
4–5 years
Underweight
6–7 years
Overweight/obese
Underweight
10–11 years
Overweight/obese
Overweight/obese
0.8
0.8
0.7
1.0
Bivariate
Disengaged (more rigid boundaries)
1.5
Enmeshed(a) (more
–
diffuse boundaries)
Number of observations
1.2
1.7
–
0.2
294
0.5
0.4
0.8
1.2
295
358
333
Multivariate
Disengaged (more rigid boundaries)
Enmeshed(a)
(more diffuse boundaries)
Regional or rural area
SEP: bottom quartile
2.8
1.9
☥
☥
2.0
0.8
☥
☥
0.2
0.4
0.7
0.9
0.7
0.3
0.6
2.0
1.2
0.9
1.1
0.8
0.7
0.7
2.0
0.7
1.3
☥
☥
1.0
Mother under 25 at first birth
1.2
1.3
0.3
0.9
0.8
1.3
0.7
1 older sibling
0.4
0.8
0.5
0.9
7.8*
0.5
1.4
2 or more older siblings
0.2
0.7
0.4
0.8
12.9*
0.6
1.5
1 or more younger siblings
0.9
0.8
1.9
1.6
5.6*
0.5
0.8
Child sees PLE 4–26 times per year
0.2
0.9
0.3
0.8
2.9
0.3**
0.6
Child sees PLE no more than twice per year
0.7
0.9
0.2
1.2
1.6
1.3
0.6
Girl
0.4
1.7
1.4
1.4
1.9
1.9
1.3
Generally healthy
0.9
1.4
0.7
1.0
1.0
0.8
0.7
0.6
2.9*
1.6
0.3**
1.3
1.1
With disability
☥
Birth weight
0.8
Indigenous
1.4
Number of observations
☥
1.8***
2.8
☥
☥
0.5*
1.1
☥
246
☥
211
☥
☥
286
(a)
Enmeshed class not extracted for families with study child 2–3 years.
Notes:
Weights applied; *** p < .001; ** p < .01; * p < .05; ☥: variable omitted due to numerical problems arising from zero or extremely low cell size.
Source: LSAC B cohort, Waves 2 and 3; K cohort, Waves 2 and 4.
37
1.3
205
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
The results in Table 6 and Table 7 highlight a number of other factors associated with
children’s weight status in families with two resident parents and families with a PLE,
respectively. In families with two resident parents, children in the bottom socioeconomic
position (SEP) quartile, across all ages, were between 1.3 and 1.7 times more likely to be
overweight/obese than in the normal weight range.
In both family types, girls 4–5 years old were significantly more likely to be
overweight/obese, but girls 10–11 years old were almost twice as likely to be underweight
than boys in the same age range. B cohort children 6–7 years old living with disability were
more likely to be underweight, and K cohort children 10–11 years old were more likely to be
overweight/obese.
A more consistent result across age groups related to children’s general health. Better health
was negatively associated with being underweight across all age groups, and with being
overweight/obese for children aged 6–7 and 10–11.
Finally, birth weight was significantly associated with children’s weight from the early years
through to middle childhood. Specifically, greater birth weight was negatively associated with
being underweight across all age groups, and positively associated with the likelihood of
being overweight/obese for all age groups except children aged10–11.
Injuries
This section presents results relating to the proportion of children injured one or more times
during the previous 12 months. Figure 13 shows the proportion of children in families with
two resident parents, and Figure 14 for families with a PLE. The proportion of children with
one or more injuries in families with two resident parents ranges from approximately 13 per
cent for children 6–7 years old across all family types to 24 per cent for children 2–3 years old
in families located toward the disengaged end of the boundary range.
Figure 13: Average percentage of children with one or more injuries in the past year in different
family groups: children 2–3 years, 4–5 years, 6–7 years and 10–11 years: families with
two resident parents
38
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Source: LSAC B cohort, Waves 2 and 4; K cohort, Waves 1 and 4.
In families with a PLE (Figure 14), rates of one or more injuries in the past year were highest
for children 2–3 years old (23–25 per cent) and again for children 10–11 years old (18–22 per
cent). The errors (and associated confidence intervals) around these averages were, however,
quite substantial. The rates fell to around 12–13 per cent for children 6–7 years old. Rates of
one or more injuries over a year were very similar across family types.
Figure 14: Average percentage of children with one or more injuries in the past year in different
family groups: children 2–3 years, 6–7 years and 10–11 years: families with a PLE
Source: LSAC B cohort, Waves 2 and 3; K cohort, Waves 2 and 4.
Table 8 and Table 9 report the bivariate and multivariate results that compare children in
cohesive families with children in other types of families across different age groups, for
families with two resident parents and those with a PLE respectively. In most models, there
was no significant difference in rates of one or more injuries between children in cohesive
families and other family types, with and without controlling for other factors and for families
with two resident parents and a PLE. The exceptions are children 2–3 and 4–5 years old in
families with two resident parents located toward the disengaged end of the boundary range.
These children were significantly more likely to have had one or more injuries requiring
medical attention during the previous 12 months, and this holds when controlling for other
factors. Post-regression tests also showed that the effect for disengaged families was
significantly different from the effect for enmeshed families, suggesting that these family
types differ significantly on this outcome (p < .05).
In families with two resident parents, children 2–3 years old living in rural or regional areas
were 1.2 times more likely to have had one or more injuries in the previous year. There was
no significant association between low SEP and child injuries, with the exception of the
model for low SEP children aged 6–7 years, who were less likely to be reported as having had
one or more injuries. Children 2–3 years old in both family types, and children 4–5 years old
living in families with two resident parents and whose mother was under 25 years at the birth
of their first child, were significantly more likely to have had one or more injuries. Girls (2–3,
4–5, and 10–11 years of age) living in families with two resident parents were significantly
39
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
less likely to have been injured one or more times in a year. Indigenous children aged 10–11
years were significantly more likely to have had one or more injuries in the previous 12
months. Effects for Indigenous status were not statistically significant in families with a PLE.
Table 8:
Odds ratios from logistic regression predicting children’s injury (one or more in past 12
months): families with two resident parents
Injured one or more times in the past 12 months
2–3 years
4–5 years
6–7 years
10–11 years
Bivariate
Disengaged (more rigid boundaries)
1.5**
1.3*
1.0
0.9
Enmeshed (more diffuse boundaries)
1.0
1.1
1.0
0.8
3,439
3,306
Number of observations
3,437
3,308
Multivariate
Disengaged (more rigid boundaries)
1.5**
1.3*
1.1
0.9
Enmeshed (more diffuse boundaries)
1.0
1.2
1.0
0.8
Regional or rural area
1.2*
1.0
1.1
1.1
SEP: bottom quartile
0.9
1.2
0.6**
1.1
Mother under 25 at first birth
1.3**
1.9***
1.2
1.0
1 older sibling
1.0
0.9
0.9
1.2
2 or more older siblings
0.9
0.8
0.7*
1.0
1 younger sibling
0.9
0.8
0.8
1.2
2 or more younger siblings
–
0.8
0.7
1.0
Girl
0.8*
0.7**
0.8
0.8*
Generally healthy
1.0
1.0
1.0
0.9
With disability
1.4
–
1.1
1.3
Birth weight
1.1*
1.0
0.9
1.1
Indigenous
1.3
1.3
1.1
2.2**
3,422
3,264
Number of observations
Notes:
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC B cohort, Waves 2 and 4; K cohort, Waves 1 and 4.
40
3,355
3,257
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Table 9:
Odds ratios from logistic regression predicting children’s injury (one or more times in past
12 months): families with a PLE
Injured one or more times in the past 12 months
2–3 years
4–5 years
6–7 years
10–11 years
Bivariate
Disengaged (more rigid boundaries)
Enmeshed(a)
(more diffuse boundaries)
Number of observations
0.9
1.3
–
0.4
301
301
0.7
0.7
0.9
364
0.8
363
Multivariate
Disengaged (more rigid boundaries)
0.9
1.3
0.9
1.1
Enmeshed(a)
–
0.4
1.1
0.7
Regional or rural area
0.9
1.1
1.5
0.9
SEP: bottom quartile
1.1
1.7
0.8
0.6
Mother under 25 at first birth
2.9**
1.5
1.0
2.2
1 older sibling
1.0
1.1
0.6
0.6
2 or more older siblings
1.3
0.7
0.2*
0.8
1 or more younger siblings
0.1**
1.3
1.1
1.4
Child sees PLE 4–26 times per year
1.3
0.4
1.4
2.6
Child sees PLE no more than twice per year
1.1
0.3
1.1
2.4
Girl
0.8
0.9
1.5
0.7
Generally healthy
1.0
1.5
0.9
0.8
With disability
3.1
1.3
1.0
0.6
Birth weight
1.4*
0.8
1.1
1.1
(more diffuse boundaries)
Indigenous
Number of observations
0.4
3.3
1.9
251
215
291
(a)
Enmeshed class not extracted for families with study child 2–3 years.
Notes:
Weights applied; *** p < .001; ** p < .01; * p < .05.
1.5
219
Source: LSAC B cohort, Waves 2 and 3; K cohort, Waves 2 and 4.
5.2
Children’s social and emotional wellbeing
SDQ prosocial score
This section considers links between the family environment and indicators of children’s
social and emotional wellbeing. Children’s social and emotional wellbeing was measured
using the Strengths and Difficulties Questionnaire (SDQ) total problem and prosocial scores
(Goodman 1997). This instrument was chosen as a key headline indicator to monitor
children’s wellbeing (see AIHW 2012b for the rationale for this choice). Figure 15 and Figure
16 show, for families with two resident parent and families with a PLE respectively, the
average SDQ prosocial score for children aged 4–5, 6–7 and 10–11 years in different family
groups. In families with two resident parents, children in more cohesive families had the
highest prosocial scores across all age groups. Children 4–5 years old in families located
toward the enmeshed and disengaged ends of the boundary range had lower prosocial scores.
Average prosocial scores were higher for children 6–7 and 10–11 years old in more enmeshed
families, but remain largely unchanged for children in families lying toward the disengaged
end of the boundary range.
41
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Figure 15: Average SDQ prosocial score in different family groups: children 4–5 years, 6–7 years
and 10–11 years: families with two resident parents
Source: LSAC B cohort, Wave 4; K cohort, Waves 1 and 4.
Figure 16: Average SDQ prosocial score in different family groups: children 4–5 years, 6–7 years
and 10–11 years: families with a PLE
Source: LSAC B cohort, Wave 3; K cohort, Waves 2 and 4.
In families with a PLE, children 4–5 and 6–7 years old in more cohesive families had the
highest average SDQ prosocial scores, and children of all ages in families located toward the
disengaged end of the boundary range consistently had the lowest prosocial scores. Children
42
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
in families located toward the enmeshed end of the boundary range lie between these two
groups but catch up with children in more cohesive families by the age of 10–11. Broadly, the
patterns were very similar across these different households. Children’s average prosocial
scores in more cohesive families were consistently high, and in more disengaged families
they were consistently low. Average prosocial scores were higher across increasing age
groups for children in families positioned toward the enmeshed end of the boundary range.
They remained somewhat different for children from cohesive families with two resident
parents, although they were roughly equivalent for children in families with a PLE.
Table 10 and Table 11 report, for families with two resident parents and families with a PLE
respectively, the results from a series of bivariate and multivariate regression analyses
considering associations between children’s SDQ prosocial scores and the family
environment, both with and without other control variables. Children in families with two
resident parents located toward the disengaged and enmeshed ends of the boundary range
have significantly lower SDQ prosocial scores than children in cohesive families, and these
differences remain after controlling for other factors (see Table 10). The difference was
smaller for children 6–7 and 10–11 years old in enmeshed families, but greater for children of
these ages in disengaged families, but the gap remains significant throughout. Post-regression
tests reveal that the effect for families toward the disengaged end of the boundary range is
significantly larger than the effect for more enmeshed families (p < .001).
Table 10:
Children’s SDQ prosocial score: families with two resident parents (OLS coefficients)
SDQ prosocial
4–5 years
6–7 years
10–11 years
Bivariate
Disengaged (more rigid boundaries)
–1.2***
–1.2***
–1.4***
Enmeshed (more diffuse boundaries)
–0.6***
–0.6***
–0.5***
Intercept
8.1***
8.6***
8.8***
Number of observations
3,302
3,418
3,273
Multivariate
Disengaged (more rigid boundaries)
–1.2***
–1.1***
–1.3***
Enmeshed (more diffuse boundaries)
–0.6***
–0.5***
–0.4***
Regional or rural area
0.0
0.1
0.0
SEP: bottom quartile
–0.1
–0.2*
0.0
Mother under 25 at first birth
0.0
0.0
0.0
1 older sibling
0.0
–0.1
–0.1
2 or more older siblings
–0.2*
–0.2
0.0
1 younger sibling
0.0
0.0
–0.2*
2 or more younger siblings
0.0
0.0
–0.3***
Girl
0.4***
0.8***
0.6***
Generally healthy
0.2***
0.2***
0.2***
With disability
0.0
–0.2
0.0
Birth weight
0.0
0.0
0.0
Indigenous
0.0
–0.1
–0.2
Intercept
7.3***
7.4***
7.9***
R-square
.11
.13
.14
Number of observations
3,260
3,336
3,224
Notes:
Weights applied; *** p < .001; ** p < .01; * p < .05.
43
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Source: LSAC B cohort, Wave 4; K cohort, Waves 1 and 4.
In families with a PLE, children of all ages in disengaged families have significantly lower
SDQ prosocial scores than children in cohesive families. However, a significant difference
between children in cohesive and enmeshed families emerged only for children aged 6–7
years, and this result held when controlling for other factors. The magnitude of the difference
for children aged 4–5 was comparable, but the standard error was greater (note the wider
confidence interval in Figure 15). As with two-parent households, post-regression tests reveal
that the effect for families toward the disengaged end of the boundary range was significantly
larger than the effect for more enmeshed families (p < .05).
Table 11:
Children’s SDQ prosocial score: families with a PLE (OLS coefficients)
SDQ prosocial
4–5 years
6–7 years
10–11 years
Bivariate
Disengaged (more rigid boundaries)
–2.0***
–1.0***
–1.0***
Enmeshed (more diffuse boundaries)
–0.6
–0.5*
–0.2
Intercept
7.8***
Number of observations
243
8.6***
350
8.7***
360
Multivariate
Disengaged (more rigid boundaries)
–1.8**
–1.2***
–1.0***
Enmeshed (more diffuse boundaries)
–0.3
–0.6**
0.2
Regional or rural area
–0.2
–0.1
0.1
SEP: bottom quartile
–0.6*
–0.1
–0.2
Mother under 25 at first birth
–0.3
–0.2
0.0
1 older sibling
–0.4
–0.1
–0.2
2 or more older siblings
–0.7*
–0.5
0.2
1 or more younger siblings
–0.2
0.1
–0.6*
Child sees PLE 4–26 times per year
0.0
–0.1
–0.2
Child sees PLE no more than twice per year
0.1
0.1
–0.2
Girl
0.5
0.4
0.5*
Generally healthy
0.2
0.0
0.5***
With disability
0.4
–0.1
0.4
Birth weight
–0.1
0.1
–0.1
Indigenous
–0.2
–0.4
–0.7
Intercept
7.4***
8.9***
6.8***
R-square
.23
.13
.29
Number of observations
179
Notes:
282
218
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC B cohort, Wave 3; K cohort, Waves 2 and 4.
Children in families with two resident parents have significantly higher prosocial scores
across all ages, and there was a positive association between health and prosocial scores (see
Table 10). Children aged 4–5 years in relatively more disadvantaged families (bottom SEP
quartile) have lower prosocial scores, but this result was not evident for older children. The
presence of siblings was also associated with lower prosocial scores for children aged 4–5
(older siblings) and children aged 10–11 (younger siblings). Girls, and children with better
health, average significantly higher prosocial scores in the models for children aged 10–11.
The coefficients were positive in all models, but not statistically significant. Siblings also had
44
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
a negative effect on prosocial scores, though this was only significant for children 6–7 years
old (older siblings).
SDQ total problems score
Figure 17 and Figure 18 show, for families with two resident parents and families with a PLE
respectively, the average SDQ total problem score for children 4–5, 6–7 and 10–11 years of
age in different family environments. At all ages, total problems were lowest among children
in cohesive families. Children in families toward the enmeshed end of the boundary range had
higher average total problem scores than children in cohesive families, but lower problem
scores than children in families positioned toward the disengaged end of the family boundary
range. At age 4–5 years, children in families exhibiting enmeshed patterns (above-average
parental conflict and average levels of parental warmth) had the highest SDQ total problem
score. However, among children 6–7 years old, those in disengaged families had higher
average total problem scores than those in enmeshed families. This gap widened further as
total problem scores for children 10–11 years old were higher in disengaged families, but
lower for children in families closely resembling enmeshed families.
Figure 17: Average SDQ total problem score in different family groups: children 4–5 years, 6–7
years and 10–11 years: families with two resident parents
Source: LSAC B cohort, Wave 4; K cohort, Waves 1 and 4.
Patterns were notably different in families with a PLE. At 4–5 years of age, children in
families located toward the centre of the boundary range (cohesive families), and those
located toward the enmeshed end of the boundary range, had the lowest average total SDQ
problem scores, and children in families positioned toward the disengaged end of this
spectrum have the highest SDQ problem scores. There was an increase in average scores for
children aged 6–7 in more enmeshed families, and an increase in average scores for children
aged 6–7 in more disengaged families. Children aged 6–7 in more disengaged families were
located below children in more enmeshed families and above those in relatively cohesive
families. This pattern is more pronounced when children were 10–11 years, as the average
45
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
total problem scores in disengaged families was higher, whereas it was lower in both
enmeshed and cohesive families.
Figure 18: Average SDQ total problem score in different family groups: children 4–5 years, 6–7
years and 10–11 years: families with a PLE
Source: LSAC B cohort, Wave 3; K cohort, Waves 2 and 4.
Table 12 shows, for families with two resident parents, the average SDQ total problem scores
for children aged 4–5, 6–7 and 10–11 years in different family groups. The results show that
children living with two resident parents in relatively more disengaged and enmeshed families
average significantly higher total problem scores than children in relatively more cohesive
families, and these differences remain significant after controlling for other factors. As was
the case with prosocial scores, post-regression tests revealed that the effect for families
toward the disengaged end of the boundary range is significantly larger than the effect for
more enmeshed families (p < .001).
Children aged 4–5, 6–7 and 10–11 years in relatively disadvantaged families (those in the
bottom SEP quartile) had significantly higher average total problem scores across all age
groups than children in relatively more advantaged families (middle and upper SEP quartiles).
Children aged 4–5 and 6–7 years in families with two or more older siblings have lower
average total problem scores than those with no older siblings.
Children 6–7 and 10–11 years old living with a disability had significantly higher average
total problem scores than those who did not have a disability (there was no equivalent
measure for children 4–5 years old). Good health was negatively associated with high total
problem scores across all age groups. Girls of all ages had lower total problem scores than did
boys, and Indigenous children 4–5 and 10–11 years of age had higher total problem scores
than non-Indigenous children of the same age.
46
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Table 12:
OLS coefficients predicting children’s SDQ total problem score: families with two resident
parents
SDQ total problems
4–5 years
6–7 years
10–11 years
Bivariate
Disengaged (more rigid boundaries)
4.2***
3.7***
6.6***
Enmeshed (more diffuse boundaries)
2.3***
2.5***
2.6***
Intercept
7.7***
7.1***
6.5***
3,418
3,273
Number of observations
3,301
Multivariate
Disengaged (more rigid boundaries)
3.9***
3.3***
6.1***
Enmeshed (more diffuse boundaries)
2.1***
2.2***
2.2***
Regional or rural area
0.0
0.2
0.2
SEP: bottom quartile
1.9***
0.9***
1.2***
Mother under 25 at first birth
1.4***
0.8***
1.0***
1 older sibling
–0.7**
–0.5*
–0.2
2 or more older siblings
–1.0***
–1.4***
–0.1
1 younger sibling
–0.1
–0.2
–0.1
0.5
–0.5
0.0
2 or more younger siblings
Girl
–0.9***
–1.2***
–1.3***
Generally healthy
–1.1***
–1.2***
–1.3***
1.5***
1.3*
With disability
Birth weight
Indigenous
–
–0.2*
–0.1
–0.1
1.9**
0.8
1.1
Intercept
12.9***
13.1***
12.3***
R-square
.21
.16
.27
Number of observations
3,259
3,336
3,224
Notes:
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC B cohort, Wave 4; K cohort, Waves 1 and 4.
Table 13 shows similar results for children in families with a PLE. Children of all ages in
disengaged families average higher total problem scores, though this was not significant for
children aged 10–11 when controlling for other factors. Children 6–7 and 10–11 years old in
enmeshed families average higher total problem scores, with and without controlling for other
factors. Post-regression tests reveal that the effect for families toward the disengaged end of
the boundary range is significantly larger than the effect for enmeshed families (p < .05).
Children 4–5 years old in relatively more disadvantaged families (those in the bottom SEP
quartile) averaged higher total problem scores than children in families in the middle or upper
SEP quartiles. Good health was negatively associated with total problem scores. There were
some other significant results relating to children of specific ages. Younger children (4–5
years old) living in rural or regional areas averaged lower total problem scores. Girls aged 6–7
years averaged significantly lower total problem scores than boys (the coefficients were of a
similar magnitude for girls aged 4–5 and 10–11 years). Lastly, children aged 10–11 years
with two or more older siblings average lower total problem scores than their counterparts
who had no older siblings.
47
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Table 13:
OLS coefficients predicting children’s SDQ total problem score: families with a PLE
SDQ total problem score
4–5 years
6–7 years
10–11 years
Bivariate
Disengaged (more rigid boundaries)
Enmeshed (more diffuse boundaries)
Intercept
2.8*
2.3**
2.4**
0.0
2.0**
3.1**
10.1***
8.7***
9.2***
241
350
360
Number of observations
Multivariate
Disengaged (more rigid boundaries)
Enmeshed (more diffuse boundaries)
3.5**
2.5**
2.1*
1.1
2.3**
2.1
Regional or rural area
–2.0**
0.1
0.4
SEP: bottom quartile
2.4*
1.3
–0.5
Mother under 25 at first birth
1.0
–0.5
1.2
1 older sibling
0.6
1.0
–1.0
2 or more older siblings
0.6
0.3
–2.6*
1 or more younger siblings
0.5
0.0
0.4
–1.6*
0.1
–0.9
Child sees PLE 4–26 times per year
Child sees PLE no more than twice per year
0.8
–0.8
–0.9
Girl
–1.6
–1.5*
–0.7
Generally healthy
–1.1*
–1.1*
–2.8***
0.9
1.6
0.1
Birth weight
–0.3
–0.1
–0.1
Indigenous
–0.8
0.8
0.4
Intercept
13.0***
13.1***
21.3***
R-square
.23
.15
.26
With disability
Number of observations
Notes:
178
282
218
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC B cohort, Wave 4; K cohort, Waves 1 and 4.
5.3
Children’s cognitive development
In this section, we present the results from analyses exploring associations between the family
environment and children’s NAPLAN reading and numeracy scores. The concurrent wave for
the majority of Year 3 NAPLAN scores is Wave 3, and for Year 5 it is Wave 4. Therefore, we
estimated models of the family environment for Wave 4 of the K cohort, when the study child
was 8–9 years old, so that we would have a more contemporaneous measure of the family
environment for use in these analyses. As in other waves, the latent class models extracted
three groups that exhibited patterns resembling those of families located toward the centre
(cohesive) or toward the more disengaged or enmeshed ends of the boundary range (results
not shown).
NAPLAN reading
Figure 19 and Figure 20 show the average NAPLAN Year 3 and Year 5 reading scores for
children in different family groups, in families with two resident parents and with a PLE
respectively. The results show an increase in NAPLAN reading scores in Year 5 compared
with Year 3. This reflects the construction of the measure, which is designed to reflect
48
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
increases in children’s abilities as they grow older. NAPLAN reading scores are very similar
for children in different family environments across all ages, though the average Year 5 score
for children in families toward the disengaged end of the boundary range drops below that for
children in other family groups.
Figure 19: Average Year 3 and Year 5 NAPLAN reading scores in different family groups:
children 8.5 years and 10.5 years (average age): families with two resident parents
Source: LSAC K cohort, Waves 3 and 4.
Figure 20: Average Year 3 and Year 5 NAPLAN reading scores in different family groups:
children 8.5 years and 10.5 years (average age): families with a PLE
Source: LSAC K cohort, Waves 3 and 4.
49
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Table 14 and Table 15 show the results from bivariate and multivariate regression analysis of
children’s Year 3 and Year 5 NAPLAN reading scores, for those in families with two resident
parents and those with a PLE respectively. Results confirm that children living with two
resident parents in more disengaged families had significantly lower average NAPLAN
Year 5 reading scores than those in more cohesive families (see Table 14). This result
remained, whether or not we controlled for other factors. Post-regression tests show that the
effect for the disengaged group was significantly greater than the effect for the enmeshed
group (p < .01)
Table 14 shows that, in families with two resident parents, children in rural and regional areas
averaged significantly lower Year 3 NAPLAN reading scores. Children in relatively more
disadvantaged families (those in the bottom SEP quartile) averaged lower Year 3 and Year 5
NAPLAN reading scores. In addition, there was a negative association between Year 3 and
Year 5 NAPLAN reading scores and having two or more older siblings. Girls averaged higher
Year 3 and Year 5 NAPLAN reading scores, and age was positively associated with
NAPLAN reading scores. There was a significant Indigenous gap for Year 5 NAPLAN
reading scores only. Children living with disability averaged significantly lower NAPLAN
reading scores at Year 3, but not at Year 5.
50
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Table 14:
OLS coefficients predicting children’s NAPLAN reading score: families with two resident
parents
NAPLAN Reading
Year 3
Year 5
(Family environment at 8–9 years) (Family environment at 10–11 years)
Bivariate
Disengaged (more rigid boundaries)
–10.7
Enmeshed (more diffuse boundaries)
–17.8***
3.8
–4.5
Intercept
429.8***
505.4***
Number of observations
2,178
2,897
Multivariate
Disengaged (more rigid boundaries)
–4.1
Enmeshed (more diffuse boundaries)
–16.0***
6.2
0.5
Regional or rural area
–10.5*
–12.7***
SEP: bottom quartile
–38.5***
–35.2***
Mother under 25 at first birth
–15.5**
–26.9***
1 older sibling
–15.5*
2 or more older siblings
–33.5***
–16.0**
1 younger sibling
1.9
4.4
2 or more younger siblings
3.1
1.9
Girl
–6.8
18.0***
24.4***
Child age at test (months)
2.1***
0.8*
Generally healthy
5.5
1.9
55.8***
–6.1
With disability
Birth weight
3.9*
0.9
Indigenous
–33.2
–48.4***
Intercept
364.4***
511.2***
R-square
Number of observations
Notes:
.12
.14
2,155
2,852
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC K cohort, Waves 3 and 4.
Results in Table 15 echoed those for children living in families with two resident parents,
though they did not reach satisfactory statistical significance (most likely due to the relatively
small sample size). There was also a negative association between Year 5 NAPLAN reading
and having one or more younger siblings. Girls averaged higher NAPLAN reading scores
than boys.
51
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Table 15:
OLS coefficients predicting children’s NAPLAN reading score: families with a PLE
NAPLAN Reading
Year 3
Year 5
(Family environment at 8–9
(Family environment at 10–11
years)
years)
Bivariate
Disengaged (more rigid boundaries)
–5.4
–2.2
Enmeshed (more diffuse boundaries)
14.1
6.0
Intercept
397.6***
481.7***
Number of observations
221
293
Multivariate
Disengaged (more rigid boundaries)
5.6
8.9
Enmeshed (more diffuse boundaries)
30.1
3.3
Regional or rural area
–4.8
–21.6
SEP: bottom quartile
14.1
–20.0
Mother under 25 at first birth
–14.2
2.4
1 older sibling
–22.2
–19.8
2 or more older siblings
–24.0
–21.5
1 or more younger siblings
–11.7
–37.1**
Child sees PLE 4–26 times per year
Child sees PLE no more than twice per
year
–30.1
–11.5
–26.5
–20.7
Girl
Child age at test (months)
Generally healthy
31.7*
29.8*
1.8
–0.9
4.6
6.6
–83.7***
12.0
Birth weight
0.2
–0.4
Indigenous
–48.2
1.1
Intercept
401.4***
With disability
R-square
Number of observations
Notes:
.20
150
496.0***
.09
183
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC K cohort, Waves 3 and 4.
NAPLAN numeracy
In this section, we consider NAPLAN numeracy scores. Figure 21 shows, for children in families with
two resident parents, the average Year 3 and Year 5 NAPLAN numeracy scores for children in
different family environments. Figure 22 provides the same information for children in families with a
PLE. The results for numeracy were very similar to the results for reading. Scores for children in
families with two resident parents were very similar across different family groups, with those in
families toward the disengaged end of the boundary range averaging lower Year 5 NAPLAN
numeracy scores. In families with a PLE, average NAPLAN numeracy scores were also very similar
for children across different family groups.
52
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
Figure 21: Average Year 3 and Year 5 NAPLAN numeracy scores in different family groups:
children 8.5 years and 10.5 years (average age): families with two resident parents
Source: LSAC K cohort, Waves 3 and 4.
Figure 22: Average Year 3 and Year 5 NAPLAN numeracy scores in different family groups:
children 8.5 years and 10.5 years (average age): families with a PLE
Source: LSAC K cohort, Waves 3 and 4.
Table 16 confirms, for children with two resident parents, that there was a significant negative
difference in Year 5 NAPLAN numeracy scores between children in more disengaged
families and those in more cohesive families, which remained when controlling for other
factors. As with reading, post-regression tests show that the effect for the disengaged group
53
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
was significantly greater than the effect for the enmeshed group (p < .01). There were strong
associations between children’s NAPLAN numeracy scores and the relative socioeconomic
position of the family.
Table 16:
OLS coefficients predicting children’s NAPLAN numeracy score: families with two resident
parents
NAPLAN numeracy
Year 3
Year 5
(Family environment
(Family environment
at 8–9 years)
at 10–11 years)
Bivariate
Disengaged (more rigid boundaries)
–2.2
–13.7**
Enmeshed (more diffuse boundaries)
2.3
1.4
422.5***
504.4***
2,179
2,894
Intercept
Number of observations
Multivariate
Disengaged (more rigid boundaries)
2.2
–13.6**
Enmeshed (more diffuse boundaries)
3.9
4.1
Regional or rural area
–11.8***
–12.9***
SEP: bottom quartile
–31.0***
–30.3***
Mother under 25 at first birth
–4.0
–27.1***
1 older sibling
–9.4
–0.4
–22.3**
–7.4
–0.1
5.5
4.6
4.1
–4.6
–0.8
2 or more older siblings
1 younger sibling
2 or more younger siblings
Girl
Child age at test (months)
1.6**
1.3***
Generally healthy
6.2*
1.3
With disability
53.2***
–10.3
Birth weight
4.0**
Indigenous
–43.2**
–43.2***
Intercept
361.2***
519.5***
R-square
Number of observations
Notes:
2.2
.10
.12
2,156
2,849
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC K cohort, Waves 3 and 4.
Children in families in the bottom SEP quartile average significantly lower Year 3 and Year 5
NAPLAN numeracy scores. There was a positive association between age and NAPLAN
numeracy scores, but there was no significant difference between girls and boys (although
there was for reading scores). Indigenous children had lower Year 3 and Year 5 NAPLAN
numeracy scores than non-Indigenous children. There was a negative association between
larger numbers of older siblings (two or more) and NAPLAN numeracy scores, but this was
only significant for the Year 5 score. Finally, as with reading scores, children living with
disability had significantly lower NAPLAN numeracy scores at Year 3 but not at Year 5.6
Table 17 shows that there were very few significant results in the models for numeracy in
families with a PLE, most likely due to the relatively small sample size. Importantly, the
results relating to family environment did not alter substantially when we controlled for other
54
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
factors. Like children living with two resident parents, children in families with a PLE living
in rural or regional areas averaged significantly lower NAPLAN Year 5 numeracy scores than
their counterparts in metropolitan areas. In addition, age was positively correlated with Year 5
NAPLAN numeracy scores.
Table 17:
OLS coefficients predicting children’s NAPLAN numeracy score: families with a PLE
NAPLAN numeracy
Year 3
Year 5
(Family environment
(Family environment
at 8–9 years)
at 10–11 years)
Bivariate
Disengaged (more rigid boundaries)
–0.7
9.7
Enmeshed (more diffuse boundaries)
6.3
2.9
Intercept
391.6***
474.1***
Number of observations
222
294
Multivariate
Disengaged (more rigid boundaries)
–9.1
8.6
Enmeshed (more diffuse boundaries)
11.8
–2.7
5.0
–20.6*
Regional or rural area
SEP: bottom quartile
8.2
–20.4
–16.2
–6.9
1 older sibling
–8.7
–15.9
2 or more older siblings
–4.1
–8.6
Mother under 25 at first birth
1 or more younger siblings
3.6
–6.8
Child sees PLE 4–26 times per year
–5.0
11.1
Child sees PLE no more than twice per year
–7.3
–3.5
Girl
12.5
–4.4
Child age at test (months)
Generally healthy
With disability
1.0
1.8*
–0.1
4.9
–47.2*
–16.0
Birth weight
–1.4
–1.3
Indigenous
–34.9
–24.8
Intercept
400.4***
496.7***
R-square
.16
Number of observations
Notes:
151
.13
184
Weights applied; *** p < .001; ** p < .01; * p < .05.
Source: LSAC K cohort, Waves 3 and 4.
5.4
Summary
In this chapter we have presented the results from analyses of concurrent associations
between family environment and children’s outcomes. There were very few significant
associations between family environment and children’s health outcomes. Children aged 2–3
years in families with two resident parents lying toward the enmeshed end of the boundary
range were significantly more likely to be underweight (than normal weight). Family
environment was not associated with other weight problems for children at other ages.
Children aged 2–3 years in families with two resident parents located toward the disengaged
55
FAMILY ENVIRONMENTS AND CHILD OUTCOMES
end of the boundary range were significantly more likely to have two or more injuries per
year.
Family environment was most strongly associated with children’s social and emotional
wellbeing. In families with two resident parents, children in families positioned toward the
disengaged end of the boundary range had significantly lower levels of prosocial behaviour,
higher levels of total problem behaviour, and higher levels of externalising problem behaviour
than children in more cohesive families. Results were very similar for children in families
with a PLE. There were also significant associations suggesting negative social and emotional
outcomes for children in enmeshed families, but these were not as pronounced compared with
the results for more disengaged families.
There were less consistent and fewer significant associations between family environment and
children’s cognitive development. In families with two resident parents, children in families
located toward the disengaged end of the boundary range averaged lower Year 3 and Year 5
NAPLAN reading scores. However, the Year 3 result was not significant after controlling for
other factors. Children in families situated toward the disengaged end of the boundary range
averaged lower Year 5 NAPLAN numeracy scores, even after controlling for other factors.
Patterns were very similar across family environments for children in families with a PLE.
These analyses were cross-sectional, and the next chapter adds to our understanding of the
relationship between family environment and child outcomes, by exploring associations
between changes in the family environment and changes in child outcomes.
56
TRANSITIONS IN THE FAMILY ENVIRONMENT AND CHILD OUTCOMES
6 Transitions in the family environment and child outcomes
This chapter addresses Research Question 5 (‘To what extent do changes in family
environment influence changes in child outcomes?’) and presents an analysis of transitions
over time. The first part of the analysis describes the nature of the transitions over time in the
B cohort and the K cohort. In families with two resident parents, we consider transitions for
families with study children aged 2–3 and 6–7 years (B cohort), and transitions for families
with study children aged 4–5 and 10–11 years (K cohort). In families with a parent living
elsewhere (PLE), we consider transitions for families with study children aged 2–3 and 4–5
years (B cohort), and transitions for families with study children aged 6–7 and 10–11 years
(K cohort). We present the results of this descriptive analysis in Section 6.1. In addition, we
explore associations between family characteristics and different transitions, using the same
variables detailed in Section 4.3. The second part of the analysis (Section 6.2) considers links
between transitions in the family environment and changes in child outcomes, controlling for
a range of factors identical to those outlined in Chapter 5.
Box 3: Chapter 6 key findings
6.1

This chapter examined changes in the family environment, and associations between these changes and
children’s outcomes. We looked at changes in the following two-year periods: 2–3 to 6–7 years
(B cohort); 2–3 to 4–5 years (B cohort, families with a PLE); 4–5 to 10–11 years (K cohort); 6–7 to 10–
11 years (K cohort, families with a PLE).

Around 54–60 per cent of families with two resident parents remained cohesive; in families with a PLE
62 per cent of the B cohort and 22 per cent of the K cohort remained cohesive.

In families with two resident parents, the family environment of 16 per cent of B cohort children
improved (that is, it became more cohesive); the family environment of 19 per cent of K cohort children
improved. In families with a PLE, the family environment of 15 per cent of B-cohort children improved
and the family environment of 20 per cent of K-cohort children improved.

Children in regional or rural areas were significantly less likely to experience a worsening of their
family environment; children with two or more siblings were significantly more likely to experience a
worsening of their family environment.

In families with two resident parents, changing family environments were significantly associated with
changes in children’s social and emotional wellbeing.

Children whose family environment improved (that is, it became more cohesive) showed improved
social and emotional wellbeing. In contrast, children whose family environment worsened (that is, it
became significantly less cohesive) exhibited increased social and emotional problems.

There were fewer significant associations between changes in the family environment and changes in
children’s cognitive outcomes. In families with two resident parents, children 10–11 years old in
families that transitioned toward the middle of the boundary range (that is, they became more cohesive)
averaged significantly higher increases in NAPLAN reading scores. There were no significant
associations between changes in family environment and changes in children’s health outcomes.
A descriptive overview of transitions in family environments
In this section, we describe transitions in family environments, beginning with transitions in
the B cohort covering a 4-year period when the study child was aged between 2–3 years and
6–7 years. Table 18 reports a cross-tabulation of cases in each family environment at two
points in time: when children were aged 2–3 years and then 6–7 years. The lower panel of
Table 18 reports the cell proportions.
Just less than 60 per cent of families remain situated toward the centre of the boundary range
at both points in time, and a further 15 per cent were toward either end of the boundary range
(either disengaged or enmeshed) when the study child was 2–3 years but moved toward the
57
TRANSITIONS IN THE FAMILY ENVIRONMENT AND CHILD OUTCOMES
centre of the boundary range when the child was 6–7 years. Therefore, three-quarters of
families were always in, or transitioned towards, a relatively more optimal family
environment over this period. Around 12 per cent of families lay toward the centre of the
boundary range (cohesive) when the study child was 2–3 years old but moved toward the
more disengaged end of the boundary range (8 per cent) or toward the more enmeshed end of
the boundary range (4 per cent) as the child grew older. About 3 per cent of families lay
toward the disengaged end of the boundary range at both time points, and 6 per cent of
families lay toward the enmeshed end of the boundary range at both time points. The
remaining families (about 5 per cent) transitioned between the enmeshed and disengaged ends
of the boundary range.
Table 18:
Cross-tabulation of family environments: B cohort children 2–3 years and 6–7 years,
families with two resident parents
Cohesive
Family environment 6–7 years
Disengaged
Enmeshed
(more rigid
(more diffuse
boundaries)
boundaries)
Family environment 2–3 years
Total
N
Cohesive
2,025
273
135
2,433
Disengaged (more rigid boundaries)
257
106
67
430
Enmeshed (more diffuse boundaries)
275
104
197
576
2,557
483
399
3,439
Total
%
Cohesive
58.9
7.9
3.9
70.7
Disengaged (more rigid boundaries)
7.5
3.1
1.9
12.5
Enmeshed (more diffuse boundaries)
8.0
3.0
5.7
16.7
74.4
14.0
11.6
100.0
Total
Source: LSAC B cohort, Waves 2 and 4.
Transitions in the family environment at two points in time were moderately similar in the
K cohort: when children were 4–5 years and then 10–11 years old (results shown in Table 19).
Table 19:
Cross-tabulation of family environments: K cohort children 4–5 years and 10–11 years,
families with two resident parents
Cohesive
Family environment 10–11 years
Disengaged
Enmeshed
(more rigid
(more diffuse
boundaries)
boundaries)
Family environment 4–5 years
Cohesive
Total
N
1,798
130
140
2,068
Disengaged (more rigid boundaries)
405
237
119
761
Enmeshed (more diffuse boundaries)
225
67
188
480
2,428
434
447
3,309
Total
%
Cohesive
54.3
3.9
4.2
62.5
Disengaged (more rigid boundaries)
12.2
7.2
3.6
23.0
Enmeshed (more diffuse boundaries)
6.8
2.0
5.7
14.5
73.4
13.1
13.5
100.0
Total
Source: LSAC K cohort, Waves 1 and 4.
58
TRANSITIONS IN THE FAMILY ENVIRONMENT AND CHILD OUTCOMES
Around 54 per cent of families remain toward the centre of the boundary range at both points
in time, and a further 19 per cent were toward either end of the boundary range (either
disengaged or enmeshed) when the study child was 4–5 years but moved toward the centre of
the boundary range when the child was 10–11 years. Therefore, just less than three-quarters of
families either were consistently in, or moved into, a relatively more optimal family
environment over this period. Around 8 per cent of families were situated toward the centre of
the boundary range when the study child was 4–5 years and moved toward either the
disengaged end (4 per cent) or the enmeshed end (4 per cent) of the boundary range.
Somewhat more families were located toward the disengaged end of the boundary range at
both time points (7 per cent), and 6 per cent were located toward the enmeshed end of the
boundary range at both time points. The remaining families (about 6 per cent) transitioned
between the enmeshed and disengaged ends of the boundary range.
Table 20 and Table 21 show similar cross-tabulations for families with a PLE. Table 20 shows
that around 62 per cent of cases remain situated toward the centre of the boundary range at
both points in time (when the study child was 2–3 years and again when the study child was
4–5 years). A further 15 per cent of families were located toward the disengaged end of the
boundary range when the study child was 2–3 years but moved toward the centre of the
boundary range when the study child was 4–5 years (that is, these more resembled cohesive
families). Fifteen per cent of families were situated toward the centre of the boundary range
when the study child was 2–3 years but moved toward the disengaged end (4 per cent) or
moved toward the enmeshed end (11 per cent) of the boundary range when the study child
was 4–5 years.
Table 20:
Cross-tabulation of family environments: B cohort children 2–3 years and 4–5 years,
families with a PLE
Cohesive
Family environment 4–5 years
Disengaged
Enmeshed
(more rigid
(more diffuse
boundaries)
boundaries)
Family environment 2–3 years
Cohesive
Total
N
186
13
32
231
46
15
9
70
Total
232
28
41
301
Cohesive
61.8
4.3
10.6
76.7
Disengaged (more rigid boundaries)
15.3
5.0
3.0
23.3
Total
77.1
9.3
13.6
100.0
Disengaged (more rigid boundaries)
%
Source: LSAC B cohort, Waves 2 and 3.
Table 21 provides a cross-tabulation of cases in each family environment at two points in time:
when children were 6–7 years and then 10–11 years. The lower panel of Table 21 reports cell
proportions. The transition patterns here differ somewhat, largely as a result of the
distribution of families across family environments for study children aged 10–11 years
(noted above in Section 4.1). Only 22 per cent of families were toward the centre of the
boundary range resembling more cohesive family environments at both time points. Fourteen
per cent of families resembled more cohesive families (situated toward the centre of the
boundary range) when the study child was 6–7 years old and then moved toward either the
disengaged end (8 per cent) or the enmeshed end (6 per cent) of the boundary range when the
study child was aged 10–11 years. A substantial group of families lay toward the disengaged
end of the boundary range at both time points (21 per cent), while a further 12 per cent moved
59
TRANSITIONS IN THE FAMILY ENVIRONMENT AND CHILD OUTCOMES
toward the disengaged end of the boundary range when the study child was 10–11 years old,
having been situated toward the enmeshed end of the boundary range when the study child
was 6–7 years. One in 10 families was positioned toward the enmeshed end of the boundary
range at both time points.
Table 21:
Cross-tabulation of family environments: K cohort children 6–7 years and 10–11 years,
families with a PLE
Cohesive
Family environment 10–11 years
Disengaged
Enmeshed
(more rigid
(more diffuse
boundaries)
boundaries)
Family environment 6–7 years
Total
N
Cohesive
80
29
22
131
Disengaged (more rigid boundaries)
39
75
6
120
Enmeshed (more diffuse boundaries)
35
42
36
113
154
146
64
364
Total
%
Cohesive
22.0
8.0
6.0
36.0
Disengaged (more rigid boundaries)
10.7
20.6
1.6
33.0
Enmeshed (more diffuse boundaries)
9.6
11.5
9.9
31.0
42.3
40.1
17.6
100.0
Total
Source: LSAC K cohort, Waves 2 and 4.
It is clear that across both cohorts, and for families with two resident parents and families
with a PLE, there was a large degree of change in the family environment over time. We saw
in Section 4.3 that different family environments exist across a broad range of families in
society. We now consider whether certain factors were associated with the transition patterns
we have just described in families with two resident parents. We do not conduct this analysis
for families with a PLE, as the cell sizes were prohibitively small.
Table 22 and Table 23 show the results from multinomial regression analysis examining
associations between family characteristics and specific transitions in families with two
resident parents. The reference family environment is ‘always cohesive’.
Table 22:
Relative risk ratios from multinomial logistic regression predicting transition: B cohort
children 2–3 years and 6–7 years
Left
cohesive
Became
cohesive
Enmeshed
to disengaged
Disengaged
to enmeshed
Regional or rural area
0.6**
0.9
1.1
0.8
No parent Y12
1.2
1.0
0.9
1.3
Jobless household
1.6
1.3
1.4
1.9
1 older sibling
1.1
1.1
2.1**
0.5
2 or more older siblings
1.5*
1.0
1.8
0.7
1 younger sibling
2.1***
0.7**
1.5
0.5
2 or more younger siblings
2.3***
0.6*
1.4
0.9
SEIFA bottom quintile
1.3
0.7*
0.6
0.7
Factors: Wave 4
Notes:
N=3,383; *** p < .001; ** p < .01; * p < .05.
Source: LSAC B cohort, Waves 2 and 4.
60
TRANSITIONS IN THE FAMILY ENVIRONMENT AND CHILD OUTCOMES
Looking at the transitions in family environment covering a period between the study child
being aged 2–3 years and aged 6–7 years, families living in rural or regional areas were
significantly less likely to move away from a relatively central position on the boundary
range. The number of siblings was positively associated with transitions away from the centre
of the boundary range (except for families where the study child aged 6–7 years has one
younger sibling). In addition, families with a study child aged 6–7 years living in the bottom
quintile of the SEIFA index of disadvantage were significantly more likely to transition away
from the centre of the boundary range (having been cohesive when the study child was 2–3
years old). Moreover, these families were significantly less likely to move toward the centre
of the boundary range over this period.
Results in Table 23 refer to transitions in family environments at two points in time, when
children were aged 4–5 and 10–11 years. Families in rural or regional areas were less likely to
have transitioned away from the centre of the boundary range or to have transitioned from the
enmeshed end of the boundary range toward the disengaged end of the boundary range,
compared with families in metropolitan areas. Families with two or more older siblings were
less likely to move from the enmeshed end of the boundary range toward the disengaged end
of the boundary range.
Table 23:
Relative risk ratios from multinomial logistic regression predicting transition: K cohort
children 4–5 years and 10–11 years
Left
cohesive
Became
cohesive
Regional or rural area
0.7*
1.1
0.5*
0.9
No parent Y12
0.9
1.2
0.9
0.7
Jobless household
1.3
0.9
1.5
2.1
1 older sibling
1.1
1.2
0.9
1.3
2 or more older siblings
1.2
1.2
0.3*
1.4
1 younger sibling
1.3
0.9
1.3
1.2
2 or more younger siblings
1.4
1.0
2.1
1.1
SEIFA bottom quintile
1.2
1.2
1.4
1.3
Factors: Wave 4
Notes:
Enmeshed
to disengaged
Disengaged
to enmeshed
N=3,253; *** p < .001; ** p < .01; * p < .05.
Source: LSAC K cohort, Waves 1 and 4.
This section has highlighted that there was a considerable degree of change in family environments
over the periods we considered. Previous analysis (Section 4.3) showed that there were not many
socioeconomic factors associated with being in a particular family environment at a single point. The
analysis here adds to this, suggesting that not only were family environments spread across different
groups in society but that these environments changed, and that transitions were also independent of a
range of socioeconomic factors. The results show, however, that children in metropolitan areas were
more likely to experience consistently more risky environments, or were more likely to move into
more risky family environments. In addition, as with the cross-sectional analysis (Section 4.3), having
two or more siblings has a negative influence on the family environment as measured in relation to the
study child. We turn now to consider associations between transitions in the family environment and
changes in child outcomes.
6.2
Multivariate analysis of transitions and child outcomes
Addressing the final research question (‘To what extent do changes in family environment influence
changes in child outcomes?’), we now present the results of multivariate analyses exploring
associations between changes in children’s outcomes related to transitions in the family over time. We
used fixed effects panel regression methods, shown to be suitable for the analysis of transitions with
two waves of panel data (Allison 1990; Johnson 2005).7 Our analysis here was focused on outcomes
61
TRANSITIONS IN THE FAMILY ENVIRONMENT AND CHILD OUTCOMES
relating to children’s social and emotional wellbeing and their cognitive development, which we found
to be strongly associated with family environment in cross-sectional analyses in Chapter 5. Measures
of these outcomes are available at two points in time for K-cohort children only. Therefore, we
restricted the analysis to these children and we analyse changes between 4–5 years and 10–11 years of
age.
Table 24 reports the results from the multivariate regression analysis. Reference families are
those where there has not been a change in the family environment. Results for children’s
social and emotional wellbeing (SDQ columns) show that children aged 10–11 years had
higher average prosocial scores and no significantly different average problem score. Broadly,
these results point to a general improvement in children’s social and emotional wellbeing over
this period.
Table 24:
Coefficients from two-wave panel regression of indicators of children’s social and
emotional wellbeing, and cognitive development: K cohort children 4–5 years and 10–11
years
SDQ
NAPLAN
Prosocial
score
Problem
Score
Reading
Age 10–11 years
0.7***
–1.3***
74.0***
Became cohesive
0.5***
–1.5***
8.9*
–2.9
Left cohesive—disengaged
–1.0***
4.0***
8.9
–8.2
Left cohesive—enmeshed
–0.3*
0.9*
7.4
–2.3
Enmeshed to disengaged
–0.9***
2.5***
–10.5
–9.4
Disengaged to enmeshed
0.6***
–1.5**
–6.3
–13.1
Numeracy
88.1***
Number of older siblings
–0.1
0.1
–1.1
1.2
Number of younger siblings
–0.1
–0.2
–4.6
13.8*
–0.5***
2.4
2.5
Good health
0.0
Intercept
7.8***
8.8***
435.6***
428.1***
Number of observations
6,573
6,572
4,152
4,152
Notes:
*** p < .001; ** p < .01; * p < .05.
Source: LSAC K cohort, Waves 1 and 4.
Compared with children whose family environment did not change across the two time
points, children who were no longer living in a cohesive family environment had significantly
smaller increases in their average prosocial scores and higher increases in their average total
problems. Conversely, children whose family environment was not cohesive at age 4–5 but
was cohesive by age 10–11 had larger improvements in prosocial scores and further decreases
in social and emotional problems than children whose family environment did not change
across these two points in time.
There were also significant and diverse changes in children’s social and emotional wellbeing
associated with changes in different types of more problematic family environments. Children
aged 10–11 years who lived in families toward the enmeshed end of the boundary range
(when aged 8–9 years) who transitioned toward the disengaged end of the boundary range
average significantly lower increases in prosocial behaviour and significantly higher levels of
problem behaviours. Conversely, children whose family environment was characterised by
higher levels of disengagement at age 8–9 and which became more enmeshed by 10–11 years
had larger increases in their prosocial scores but unchanged problem scores.
There were fewer significant associations between changes in cognitive outcomes and
changes in the family environment (NAPLAN columns, Table 24). Children aged 10–11
62
TRANSITIONS IN THE FAMILY ENVIRONMENT AND CHILD OUTCOMES
years in families that transitioned toward the middle of the boundary range averaged
significantly higher increases in NAPLAN reading scores. However, similar effects, though
not statistically significant, were found for children in families that had transitioned from a
more cohesive family environment toward the disengaged or enmeshed ends of the boundary
range. Moreover, effects relating to these transitions and NAPLAN numeracy scores were
negative (but again not statistically significant). Together, these results paint a relatively
inconsistent picture, but broadly imply relatively weak links between these measures of the
family environment and children’s NAPLAN scores.
Table 25 reports on families with a PLE and the influence of changing family environment on
changes in children’s social and emotional wellbeing, and cognitive development from ages
6–7 to 10–11 years. As with children who have two resident parents, there was a significant
increase in children’s prosocial behaviours and a significant decrease in total problems in
families with a PLE at age 10–11 years compared with age 6–7 years.
There were fewer significant associations between changes in family environment and
changes in wellbeing for children in families with a PLE. Children in families that
transitioned from an enmeshed to a disengaged family environment across both time points
averaged significantly higher total problem scores.
Table 25:
Coefficients from two-wave panel regression of indicators of children’s social and
emotional wellbeing, and cognitive development: K cohort children 6–7 years and 10–11
years: families with a PLE
SDQ
NAPLAN
Prosocial
score
Problem
score
Reading
Numeracy
Age 10–11 years
0.8***
–1.2**
80.0***
110.8***
Became cohesive
0.3
0.3
–6.1
–32.6
Left cohesive—disengaged
0.1
0.3
–21.5
–48.2
Left cohesive—enmeshed
0.4
0.5
32.6
–11.1
Enmeshed to disengaged
–0.6
2.5*
55.9
12.7
Disengaged to enmeshed
0.1
2.8
–15.3
–10.6
Number of older siblings
0.0
–0.3
12.6
–14.4
Number of younger siblings
–0.4
1.2
26.4
3.3
General health
–0.1
0.1
3.0
17.5
380.4***
370.5***
Intercept
8.1***
10.1***
Number of observations
723
723
Notes:
526
526
*** p < .001; ** p < .01; * p < .05.
Source: LSAC K cohort, Waves 2 and 4.
6.3
Summary
This chapter has examined transitions in the family environment, and associations between
these transitions and changes in children’s outcomes. Results highlighted a considerable
degree of transition between different family environments observed at two points in time.
Around 54–60 per cent of families with two resident parents remained cohesive, 62 per cent
of B-cohort families with a PLE remained cohesive (study child aged 2–3 to 4–5 years), and
only 22 per cent of K-cohort families remained cohesive (study child aged 6–7 to 10–11
years).
63
TRANSITIONS IN THE FAMILY ENVIRONMENT AND CHILD OUTCOMES
In families with two resident parents, the family environment of 16 per cent of B-cohort
children improved (that is, it became more cohesive); the family environment of 19 per cent
of K-cohort children improved. In families with a PLE, the family environment of 15 per cent
of B cohort children improved; the family environment of 20 per cent of K-cohort children
improved.
Children in regional or rural areas were significantly less likely to experience a worsening of
their family environment; children with higher numbers of siblings were significantly more
likely to experience a worsening of their family environment.
Changing family environments were significantly associated with changes in children’s social
and emotional wellbeing. Children whose family environment improved (that is, it became
more cohesive) showed improved social and emotional wellbeing. In contrast, children whose
family environment worsened (that is, it became significantly less cohesive) exhibited
increased social and emotional problems. Associations between family environment and
children’s social and emotional outcomes were less frequent in families with a PLE. Children
whose families were characterised by high enmeshment at age 6–7 years and then by
disengagement at age 10–11 years averaged lower prosocial behaviours and higher
externalising problems. This highlights the particular difficulties associated with disengaged
families.
There were fewer significant associations between changes in the family environment and
changes in children’s cognitive outcomes. Children aged 10–11 years in families that
transitioned toward the middle of the boundary range (that is, more cohesive) averaged
significantly higher increases in NAPLAN reading scores. There were no significant
associations between changes in family environment and changes in children’s health
outcomes.
64
SUMMARY AND DISCUSSION
7 Summary and discussion
7.1
Background
Families play a crucial role in protecting and supporting children. The vast majority of
Australia’s children live, grow and thrive in safe and supportive family environments, but
increasing numbers of children have been coming to the attention of statutory child protection
services over the past couple of decades. Not only are more children becoming known to child
protection services, but also the range of problems and issues faced by these children and their
families extends beyond the most extreme forms of abuse and neglect to encompass broader
social problems and family dysfunction (Bromfield et al. 2010). This has led to a reframing of
the policy approach to child protection away from focusing mostly on statutory responses to
risk-of-harm reports (‘tertiary services’) toward a broad-based public health approach that
incorporates services targeted to those families potentially at risk (‘secondary services’),
combined with a primary prevention focus on the role of universal services to support the
broader population of all families (O’Donnell, Scott & Stanley 2008). This approach is
premised on the understanding that risks to children’s safety and wellbeing exist on a
continuum, and that protecting Australia’s children is everyone’s responsibility. As part of
this approach, Australian governments and community agencies are committed to ensuring
that children live in safe and supportive family environments and closely monitoring
indicators of children’s health, development and wellbeing. In this discussion, we review the
aims of the research, discuss the main findings and acknowledge the limitations.
7.2
Aims of the research
This research report addressed a number of themes that relate to this shift in policy focus
toward a public health approach to child protection. It is important that we understand more
about the nature and prevalence of different types of family environments across the
population, which was the first aim of this report. We assume that the majority of families do
provide safe and supportive environments for children; however, it is useful to have a more
certain understanding of how many families do, and the characteristics of these families.
Moreover, it is important to know about other types of family environments in society.
The second aim of this report was to increase our understanding of how children are faring on
a range of indicators of health, cognitive outcomes and wellbeing. Monitoring children is
central to protecting children, and appropriate data are pivotal to success. This report
highlights the potential of survey data (in particular, the Longitudinal Study of Australian
Children or LSAC) to track the safety and wellbeing of Australian children over time.
Lastly, this report sought to build knowledge around the links between family environments
and child outcomes, which could provide insights to help improve the degree to which
children’s environments promote safety and support.
7.3
Analysis
In line with these aims, this report contains three analytical components. First, using latent
class cluster analysis—a technique that has been used in previous research in this area—we
sought to identify different family groups based on factors elucidated in previous theoretical
and empirical research (parental warmth, angry parenting, parental conflict and parent–child
shared activities). We profiled different family groups and explored social and demographic
factors associated with them.
65
SUMMARY AND DISCUSSION
Second, we conducted a detailed descriptive analysis of a range of indicators of children’s
health (BMI and injuries), social and emotional wellbeing (Strengths and Difficulties
Questionnaire—prosocial and total difficulty scores), and cognitive development (NAPLAN
reading and numeracy scores). We provided a snapshot of how children were faring on these
indicators across a wide range of ages from infancy to middle childhood.
Third, we conducted a series of multivariate cross-sectional analyses exploring links between
the family environment and child outcomes concurrently. Additionally, using descriptive and
longitudinal multivariate analyses, we explored the relationship between transitions in the
family environment and changes in child outcomes. We conducted these analyses using
multiple waves of data from LSAC, for families with children ranging from infancy to middle
childhood.
7.4
Findings: identifying family environments
We identified three distinct groups of family environments with profiles (in terms of parental
warmth, angry parenting, parental conflict and parent–child shared activities) that closely
resemble different family environments articulated in previous theoretical and empirical
research. The largest group of families displayed average or above-average parental warmth
and parent–child shared activities, and below-average angry parenting and parental conflict.
The literature refers to families resembling this profile as cohesive (for example, Kerrig
1995). We suggest that these families represent an exemplar of a safe and supportive family
environment. As we would expect, these families were the majority, supporting the
proposition that most Australian children live in safe and supportive environments.
In addition to the cohesive family environment, our analyses showed two further types of
family environments. These two groups were both smaller than the group of cohesive
families—although the first of these two groups was slightly larger than the other. This group
averaged lower parental warmth and higher angry parenting. In addition, it exhibited belowaverage parent–child shared activities and average or slightly above-average parental conflict.
These families are most similar to disengaged families described in previous research (see
Minuchin 1978). In such families, there are rigid boundaries (as demonstrated by lower
parental warmth) and a tendency to close off access to resources for children.
The last group of families had strikingly higher levels of parental conflict than the other two
groups. In addition, they had average or slightly above-average levels of parental warmth and
parent–child shared activities. These patterns arise in families with boundaries that tend to be
diffuse, and these families have been referred to as enmeshed in previous research (see
Minuchin 1978). Higher levels of parental conflict that tends to negatively affect parenting
and lower levels of parent–child interactions distinguish these family environments from the
two other groups. These groups emerged clearly from our analysis across families with study
children of different ages, but there was less clarity among the families with the youngest
study child (aged 2–3 years). Previous research considered only families with school-age
children, and it may be that these dimensions of the family environment become more distinct
as children enter school.
It is important to stress that, across families, the parent factors we assessed (shared activities,
warmth, and parent–parent conflict) varied considerably across the full range of the scales we
used. In other words, in general we did not observe many families with scores indicating
extreme disengagement or enmeshment. This does not detract from the utility of the findings,
however. In fact, our results highlight that risks to children’s safety and wellbeing operate
along a continuum spanning all families. There was some limited association between
family environment and socioeconomic status. Specifically, we found that low-income
66
SUMMARY AND DISCUSSION
families with older study children aged 6–7 and 10–11 years were significantly more likely to
be in the enmeshed family group. In addition, children aged 2–3 years in relatively lesseducated households were more likely to be in more disengaged families, but this did not
extend to children older than 2–3 years. Therefore, at different points in children’s lives,
different aspects of socioeconomic status are associated with particular aspects of family
environments. There is not, in other words, a consistent pattern. This provides some support
for the validity of a public health approach to child protection, because it shows that factors
associated with risks for children are evident to a greater or lesser degree across the entire
population (as observed with nationally representative survey data). Of course, it is important
to recognise that looking at parenting behaviour and parental conflict is not the only way to
assess whether an environment is safe and supportive.
Often, statutory child protection authorities and the secondary service system (support for
families needing extra assistance, with a focus on early intervention) focus their efforts
towards low-SES families, where many of the risks of child maltreatment are congregated
either because service delivery (and surveillance) is concentrated in areas of geographic
disadvantage or because services are otherwise allocated to those with the greatest apparent
need. However, this is not to assume all children growing up in poverty have worse
outcomes—or that all socioeconomically advantaged children are doing well. Our results
suggest to some extent that potentially problematic dynamics within the families are not
concentrated in particular socioeconomic groups. In other words, our results may identify
certain problematic behaviours associated with parent–parent and parent–child relationships,
rather than specific groups in the population. This implies that the targeting of services to
those most in need could be enhanced by identification of families with problematic intrafamilial dynamics and targeting people by behaviour rather than targeting people by
demographic characteristics. Furthermore, our findings suggest that a single one-size-fits-all
approach may be unsuitable. Different family environments are likely have different needs
requiring different types of responses. This suggests that public health campaigns that address
parenting practices across the population may be an effective means of addressing the more
problematic family environments identified in this report, as population-wide screening of
parenting behaviours may not be cost-effective and may have unintended consequences.
However, existing services that come in contact with many parents (for example, perinatal
services) could have a role in identifying those with seriously problematic family dynamics.
Moreover, as well as clearly identifying distinct family environments, our results revealed a
sizeable degree of change in family environments observed between two points in time.
Transitions into a more problematic family environment in the B cohort (aged 2–3 to 6–7
years) were associated with having a larger number of siblings, which may reflect heightened
stresses for parents in these relatively early stages in family life. Transitions into a more
problematic family environment in the K cohort (aged 4–5 to 10–11 years) were more likely
in metropolitan areas. This may reflect the increasing influence of external (outside the
family) factors, as children grow older. There were not, however, any clear associations
between socioeconomic factors and changing family environments.
More generally, these transitions bring attention to the dynamic nature of family
environments, which opens up the potential for intervention to effect change (we elaborate on
possible interventions later in the discussion). They also underscore the merits of adopting a
longitudinal perspective when addressing families’ needs. It is important to recognise that,
while families may currently be experiencing difficulties, they may have had positive
experiences of family cohesion upon which they could draw. It is important to note that we
measured change between two points in time only, and it may be that there was variability in
the family environment that we did not observe in this study.
67
SUMMARY AND DISCUSSION
We carried out these analyses using data from families where the study child lived with two
resident parents, and families where the study child had a parent living in a separate
household from the primary respondent (referred to as a parent living elsewhere or PLE).
Broadly, results here were consistent across both family types. However, there were some
differences. When we looked at families with study children aged 2–3 years, there were no
families with a PLE who were characterised with high levels of enmeshment, though these
were clearly apparent in families with a PLE when study children were older (aged 4–5, 6–7
and 10–11 years). Another difference between families with a PLE and families with two
resident parents (with study children aged 6–7 and 10–11 years) was that the proportion with
profiles that most resembled cohesive families (higher parental warmth and involvement and
lower parent and anger) fell well below 50 per cent for families with a PLE (31.0 per cent at
age 6–7 and 40.1 per cent at age 10–11, cf. children at the same age in two resident parent
families: 74.4 per cent and 73.4 per cent respectively; see Tables 2 and 3 earlier). That is,
families with lower parental warmth and involvement and families with higher parental
conflict were proportionally more common. It is important to stress that the absolute numbers
of low-cohesion families with a PLE are small; however, this is a notable and statistically
significant difference. It is known that children with single or separated parents are
disproportionately known to child protection services (Hunter & Price-Robertson 2012) and
generally fare worse on a number of key indicators (McLanahan & Sandefur 1994). The
results from our current analyses suggest that, even in a relatively normative sample, key
differences in family structure relate to different family environments.
7.5
Findings: family environments and child outcomes
Having considered the prevalence and profiles of different family environments, we turned
our attention to links between these environments and child outcomes. Our results showed
little association between children’s health outcomes and their family environments. We
found that children aged 2–3 years living in families with significantly higher than average
levels of parental conflict were more likely to be underweight than normal weight. However,
as children grew older, the association between underweight and more problematic family
environments did not remain and was not evident in children in the older cohort, suggesting
that it is not a robust or lasting association.
At particular points in time, the family environment was associated with children’s cognitive
outcomes. Children in families with less-involved parents (and higher levels of angry
parenting) had significantly lower Year 5 NAPLAN reading and numeracy scores. We did not
observe these differences for Year 3 scores, however, and we did not observe significant
associations between changes in the family environment and changes in children’s cognitive
outcomes.
In contrast to health and cognitive outcomes, we found significant and consistent associations
between children’s social and emotional wellbeing and their family environment. Children in
families with higher levels of cohesion (with above-average warmth and parent–child
involvement, and lower angry parenting and parental conflict) had significantly higher levels
of prosocial behaviour and significantly lower levels of problem behaviour than children
living in other family environments. The difference is especially large when compared with
children who live in families with lower than average levels of parental warmth and parent–
child involvement (who are most similar to disengaged families described in theoretical
work).
Furthermore, we found that changes in children’s family environment were significantly
associated with changes in their social and emotional wellbeing. Children whose family
environment moved closer toward one resembling cohesive families (more warmth and
68
SUMMARY AND DISCUSSION
involvement, less anger and conflict) exhibited increased prosocial behaviour and decreased
problem behaviour. The reverse was the case if their family environment moved away from
being a more cohesive environment. Although these results relate directly to social and
emotional wellbeing, it is important to emphasise that there may be links between
socioemotional outcomes and other child wellbeing outcomes (AIHW 2011b; Hamilton &
Redmond 2010). Therefore, family environments that promote socioemotional wellbeing are
likely to have benefits for other domains of child wellbeing.
It is perhaps not surprising to find that children’s social and emotional wellbeing is most
significantly associated with their family environment measured as a function of indicators of
parent–child and parent–parent psychosocial interactions (we further elaborate on this point
below when discussing the limitations of the study). Nevertheless, these results highlight the
influence of this aspect of the family environment on this important dimension of children’s
wellbeing. The fact that we observe significant associations between changes in the family
environment and changes in children’s outcomes is a particularly strong aspect of our
findings. The particularly strong negative effects for children in families with lower parental
warmth and involvement point to the importance of the family in providing children with a
secure base and a sense of connection or togetherness (Bowlby 1988). It is important not to
forget, though, that children in families marked by higher levels of parental conflict also
exhibit relatively poorer social and emotional outcomes.
7.6
Implications for policy
We have alluded in general terms to the possible implications for policy that arise from these
findings. To summarise, with respect to identifying different family environments, we found:
1. we could identify that different family environments exist
2. they are not strongly related to factors we would normally associate with
difficulties with the family affecting child welfare (such as socioeconomic factors)
3. they are subject to change over time.
The results linking family environments to key child outcomes (especially around social and
emotional wellbeing) suggest that the efficacy of policy may be enhanced if policies and
services:
1. are attuned or sensitive to different family environments
2. target behaviour (parental family dynamics) rather than people on the basis of
their sociodemographic characteristics
3. recognise both that families can change for the better and that they can potentially
draw on their own prior (positive) experiences
4. are based on a public health approach to promote safe and supportive family
environments.
The aim of a public health approach to protecting children is to shift the focus away from a
narrow band of children requiring statutory intervention toward addressing the needs of all
families, and to move the population distribution on risk factors toward the positive end for
all families. Shifting the profile of all families would potentially reduce the number that
would be at risk of statutory intervention and improve the daily lives of many children. In
terms of public health interventions, three possibilities arise, and this study may provide some
helpful insights. The three potential types of interventions are (a) parenting programs and
supports, (b) public information programs and (c) targeted referrals for more intensive family
support.
69
SUMMARY AND DISCUSSION
Parenting programs and supports
Parenting programs have been widely used in early intervention strategies targeted toward
vulnerable families (for example, the New South Wales Government’s Brighter Futures
program and the Commonwealth Government’s Stronger Families and Communities
Strategy). However, some argue that parenting programs can be delivered as part of a public
health approach to strengthen and support parenting (Sanders 2008), and to prevent child
maltreatment (Sanders, Cann & Markie-Dadds 2003; Sanders & Pidgeon 2011). Prinz et al.
(2009) provided evidence of a significant prevention effect following from the delivery of a
parenting program in the United States. An Australian example, the Every Family initiative,
trialled the delivery of the Triple P-Positive Parenting Program in 30 sites across Brisbane,
Sydney and Melbourne (Sanders et al. 2005).8 Prinz et al. (2009) articulated a number of
criteria essential for success in a public health initiative of this nature (see also Sanders 2012).
The first three criteria referred to 1) the importance of having a good understanding of the
prevalence of the particular problem behaviours in children being targeted, 2) the importance
of knowledge pertaining to the prevalence of parent risk and protective factors, and 3)
knowledge that changing risk and protective factors improves child outcomes.9 We think that
our report provides a range of information relating to these criteria.
There is a range of other evidence-based approaches to supporting parents and addressing
problematic parenting behaviour—for example, through individual parenting education,
counselling and mediation (particularly in the context of parental separation). Parental
education and support is also a key feature of home visiting programs (see Holzer et al. 2006).
Wise et al. (2005) provided other examples of parenting supports and early childhood
interventions whose efficacy is supported by good research evidence. Mildon and Polimeni
(2012) reviewed programs that have specifically targeted Indigenous families.
A large body of research provides strong evidence that the home environment—in particular,
concrete behavioural patterns of parents (that is, parenting characteristics)—is an important
determinant of children’s early development and wellbeing. However, it should be noted that,
while parenting programs (even those with the highest evidence of their effectiveness,
particularly those that are modularised, structured, manualised et cetera) and home visiting
programs (which are usually a suite of services that may include particular components such
as parenting programs and coaching or mentoring) have been shown to improve parenting
skills, with the notable exception of Prinz et al. (2009), there is no strong evidence that they
are sufficient to prevent child maltreatment (Holzer et al. 2006; Mildon & Polimeni 2012).
Public information campaigns
Public information programs are a more familiar tool used by government to effect broader
changes in the behaviour of the population in general. Examples abound, including public
health campaigns around alcohol, smoking and skin cancer, and drink-driving and safedriving campaigns. Recent campaigns highlighting how parental alcohol consumption affects
children offer an interesting template for how such campaigns can be used to educate parents
about the influence they have over children.10 Research has explored the utility of popular
media to promote positive parenting practices more generally (Sanders & Prinz 2008) and to
promote the prevention of child maltreatment (Saunders & Goddard 2002). However, there
are limitations to public health campaigns that should be borne in mind. There is limited
evidence to address the question of whether or not social marketing campaigns are effective in
addressing concrete outcomes like rates of child abuse and neglect—and evaluations of such
campaigns are notoriously difficult to conduct with any rigour (Horsfall, Bromfield &
McDonald 2010).
70
SUMMARY AND DISCUSSION
Consistent with the World Health Organization Ottawa Charter for Health Promotion,11 a
range of actions can be taken to improve outcomes, based on advocacy, enabling people to
take control of factors that affect their wellbeing, and mediating between differing interests in
society for the pursuit of health. While public information programs can assist, there are
limitations to their effectiveness, particularly when knowledge or attitudes alone are
insufficient to effect change. They need to be targeted at attitudes or behaviours that are
modifiable, with clear links to strategies for achieving the desired change. Adopting a broad
information campaign may have limited effect if it is not directed toward behaviours that can
be changed and does not point to sources of support for bringing about that change. For
example, the national and state/territory ‘Quit’ initiatives are effective in responding to the
problem of smoking because they are targeted at broad social attitudes as well as concrete
actions and supports for quitting smoking.
Targeted referrals for more intensive family support
Often the distinction between universal and targeted services is presented as a dichotomy;
however, there is scope for it to be seen as a continuum, with universal services being the
platform for the ramping up or integration of services that would then be classified as
targeted. The principle of proportionate universalism (or progressive universalism, as it is also
termed) was outlined in the Marmot review12 of the social determinants of health inequalities
in the United Kingdom. According to this principle, actions must be ‘proportionate to the
degree of disadvantage, and hence applied in some degree to all people, rather than applied
solely to the most disadvantaged’ (Lancet 2010, p. 525). It is also important to remember that
disadvantage is not static—families (or even communities) can move into and out of
disadvantage.
While services targeted at the most disadvantaged have the greatest impact, it is also true that
targeted services would then mean the majority of the population misses out. Our analyses
have shown that less than optimal parenting practices and family environments are not
restricted to particular demographic groups and cannot be easily targeted—so there is value in
considering the role of universal services to deliver information, supports and services for all
Australian families, with increased intensity for those who need it most. Universal services
can provide the platform to refer people who require them to more specialist services, or
provide a continuum of service, so that within the universal service platform more intense
services can be provided to those in need. A number of authors have argued for the
importance of using universal services as a base or soft-entry point for engaging families that
might otherwise be hard to reach (Muir et al. 2009; O’Donnell, Scott & Stanley 2008; Scott
2006).
7.7
Limitations
This study has a number of limitations that should be borne in mind when evaluating the
findings and any implications that may follow from them. The analysis is limited in terms of
the types of measures used to identify different family environments, which correspond to a
relatively narrow range of the potential risks children face. We concentrate (following
previous theoretical and empirical work) on psychosocial aspects of the family environment
relating to parent–child and parent–parent relationships. These factors do not capture, at least
directly, important aspects of the family environment relating to learning and health. For
example, we do not measure how cognitively stimulating the family environment is, which is
likely to be related to cognitive outcomes. Moreover, we do not measure those aspects of the
family environment relating to physical health, such as diet or physical activity, likely to be
related to health outcomes. LSAC contains indicators that could be used to measure these
71
SUMMARY AND DISCUSSION
aspects of the home environment for future research. Grounded in family systems theory, the
comparative strength of our findings on the links between children’s social and emotional
wellbeing and their family environment follows from our measurement of the family
environment foregrounding psychosocial dimensions of parent–child and parent–parent
relationships. This does not detract in any way from our findings, but we should not discount
(based on this analysis) the potential influence of other dimensions of the family environment
on related child outcomes.
As well as limitations in the comprehensiveness of the assessments of the family environment
in LSAC, all of our measures are in reference to the study child. For example, the information
about warm parenting relates to the responding parents’ behaviours toward the study child
only. We found that having two or more siblings in the family was associated with lower
levels of parental warmth and parent–child shared activities (factors closely associated with
disengaged families). This may reflect a shifting of attention away from the study child onto
other children in the family, and it is possible that a whole-of-family approach to measuring
the family environment would lead to alternative interpretations. Note that our child outcomes
related specifically to the study child, and it is therefore appropriate that we adopted an
approach that is focused on the study child when measuring the family environment.
However, it is clear that dynamics within families between parents and multiple children, as
well as between siblings, are likely to have a bearing on both the family environment and
outcomes for children. Future work could explore this further.
We found significant associations between transitions in different family environments and
changes in children’s social and emotional wellbeing. However, it is important to bear in
mind that children can influence the aspects of the family environment that we consider as
well as being influenced by them (Holden 2010; Maccoby & Martin 1983). The analysis here
has not considered bi-directionality in the relationships between children’s social and
emotional outcomes and the broader psychosocial family environment. This is likely to be an
important piece of the story, and future research could examine this. Ultimately, this places
limits on the extent to which we can attribute a causal relationship between the family
environment and child outcomes. Notwithstanding this, the transition analysis we present is a
powerful indication of a strong relationship between family environment and aspects of child
outcomes.
We explored the family environment and its links with child outcomes in families with two
resident parents and in families with a parent living elsewhere from the primary respondent.
In the latter case, this report broke important new ground. Although most of the results were
similar across these families, we could trace some important differences that may have arisen
in some part due to data limitations. In particular, the measure of conflict between parents was
more limited for families with parents who do not live together. The conflict between parents
in these families may differ, and it would be important to capture this. Moreover, there is
scope, and perhaps increasing need, for further theoretical development in this area. In
addition, there were limits in the measurement of parenting for the PLE and the PLE family
environment more generally (for example, whether the parents had a new partner), which may
have limited our ability to fully capture these dimensions of the family environment. Future
work could explore the varied and complex family environments that these children
experience. Note also that we have not considered changes in family structure, as this analysis
is restricted to families where the structure (that is, families with two resident parents or with
a PLE) does not change. Future work could examine changing family structure.
Our analysis does not focus specifically on highly disadvantaged families who are more likely
to be known to the current statutory child protection and child welfare sectors in Australia.
Moreover, there is no direct measure of child maltreatment in LSAC, and the analysis in this
72
SUMMARY AND DISCUSSION
report cannot be understood as or used as a proxy for such a measure. The report is written
around the context of a public health approach to child protection, which aims to ensure that
all children grow up in safe and supportive environments. We are not attempting to identify
maltreatment or those at specific risk. Although children most at risk tend to be concentrated
in circumstances of relatively high disadvantage, a public health approach would seek to
broaden the policy focus to address wider needs that will make positive changes for the bulk
of the population. The research is intended to inform policies to address most Australian
families, so that child protection systems have to deal with a decreasing proportion of families
for whom a public health approach is not appropriate.
Although we found better socioemotional outcomes for children in families with higher levels
of cohesion, we do not know whether abuse or neglect occurs even within these family
environments. Conversely, we cannot, and do not, state that abuse or neglect is occurring in
the more problematic families we identified. Therefore, this report provides little direct
insight into the most problematic families from a child protection viewpoint. However, this
analysis does underscore the potential usefulness of a public health approach to enhancing
safe and secure family environments for children.
7.8
Conclusion
This report has shown that, in broad terms, there are distinct family environments across
society that are similar in certain factors associated with parent–child and parent–parent
interactions, and that these groups are not directly linked to particular socioeconomic groups.
The majority of families had high levels of parental warmth and involvement with children. A
smaller group, though substantial, were different in that they had below-average parental
warmth and parent–child involvement, and tended to have above-average levels of angry
parenting. A third group, equally substantial, was notable for significantly higher levels of
parental conflict but average levels of parenting warmth and parent–child involvement. We
observed these environments repeatedly over successive waves of a longitudinal dataset, but
there was considerable change in the family environment within families over time.
We found that these different family environments, with their dynamic nature, have a strong
influence on certain child outcomes, particularly those relating to children’s social and
emotional wellbeing. Children with warm, highly involved parents had higher social and
emotional wellbeing. Those with less involved parents, and who experienced above-average
angry parenting, tended to have lower social and emotional wellbeing. Children in families
marked by higher levels of parental conflict were between these two groups. This highlights
the importance of parent–child and parent–parent interactions in shaping aspects of the family
environment that children’s social and emotional wellbeing are sensitive to.
This report highlights the potential for public health interventions aimed at improving
dimensions of the family environment that are strongly associated with children’s social and
emotional wellbeing. Possible interventions include parenting programs and public
information programs. This study suggests that careful tailoring of interventions to specific
dynamics arising within families would be beneficial, and that programs must be able to reach
a broad cross-section of society.
This report emphasises the importance of families in providing children with a safe and
supportive environment. We asserted, and have found, that most do. Moreover, based on the
indicators we examined, children in relatively more problematic families are as healthy and
largely doing as well in school as other children, but there are some problems associated with
their social and emotional wellbeing. The public health space provides government with an
73
SUMMARY AND DISCUSSION
opportunity to intervene to help these families, recognising that these problems could arise in
any family at any time.
74
SUMMARY AND DISCUSSION
Appendix: Technical details about class selection
Following Sturge-Apple, Davis and Cummings (2010), we estimated a series of latent class
models extracting two, three and four classes, inspecting a number of key statistics to decide
on the optimal number of classes. Specifically, we examine the sample-adjusted Bayesian
information criterion (BIC). Lower BIC values indicate a better fitting model. In addition, we
consider the Vuong-Lo-Mendell-Rubin (VLMR; Lo-Mendell & Rubin 2001) likelihood ratio
test, which compares the k-1 model to the k model. That is, it compares whether two classes
are preferred to three, three to four and so on. When the p-value is greater than .05, the k-1
model is preferred to the k model. Note that we estimate a five-class model to extract this
statistic for the four-class solution. We were also mindful of practical considerations with
respect to the extent to which the numbers in each class were sufficient for subsequent
multivariate analyses.
Table A1 reports the sample-adjusted BIC and the VLMR p-values from latent class models
extracting two, three and four classes (the VLMR p-value for the four-class model is taken
from a five-class model). Clearly, the addition of a further class reduces the value of the
sample-adjusted BIC. However, the VLMR p-value for the four-class solution is not greater
than .05 in all cases except in the model for families with study children aged 4–5 years. In
contrast, across all years, the three-class solution is preferred to the four-class solution
(VLMR p > .05). This is strongly in line with previous research and theoretical expectations,
and we proceed with the three-class solution.
Table A1:
Sample-adjusted BIC and Vuong-Lo-Mendell-Rubin p-values: families with two resident
parents
Sample-adjusted BIC
VLMR p-value
Two classes
Three classes
Four classes
2–3 years
57989
57552
56822
4–5 years
57860
57396
6–7 years
58315
57735
10–11 years
56399
55862
Notes:
Two classes
Three classes
Four classes
.00
.22
.00
57177
.00
.24
.13
57443
.00
.38
.04
55582
.00
.69
.05
N (2–3 years and 6–7 years) = 3,439; N (4–5 years and 10–11 years) = 3,309.
Source: LSAC B cohort, Waves 2 and 4; K cohort, Waves 1 and 4.
We repeated this process for families with a parent living elsewhere (PLE). We estimated
models for families with a study child 2–3 years (B cohort, Wave 2), 4–5 years (B cohort,
Wave 3), 6–7 years (K cohort, Wave 2) and 10–11 years (K cohort, Wave 4). Table A2
reports the sample-adjusted BIC and the VLMR p-values from latent class models extracting
two, three and four classes (the VLMR p-value for the four-class model is taken from a fiveclass model). The sample-adjusted BIC decreases with increasing numbers of classes, but the
VLMR p-values are greater than .05 for all options.
75
SUMMARY AND DISCUSSION
Table A2: Sample-adjusted BIC and Vuong-Lo-Mendell-Rubin p-values: families with a PLE
Sample-adjusted BIC
Two classes
Three classes
VLMR p-value
Four classes
Two classes
Three classes
Four classes
2–3 years
2528
2503
2503
.15
.80
.85
4–5 years
3438
3430
3357
.33
.80
.33
6–7 years
2507
2481
2471
.19
.86
.25
10–11 years
3531
3505
3483
.17
.77
.77
Notes:
N (2–3 years and 4–5 years) = 301; N (6–7 years and 10–11 years) = 364.
Source: LSAC B cohort, Waves 2 and 3; K cohort, Waves 2 and 4.
The four-class solution yielded one group with a very small number of cases, which would be
of little value in subsequent analyses. As there was strong theoretical support for a three-class
solution, in cases where the number of cases in the smallest class was sufficient for
subsequent analysis, we chose the three-class solution. In families with a study child aged 2–3
years, the three-class solution yielded a group with only 14 cases, and we therefore opted for
the two-class solution.
76
List of shortened forms
AIHW
Australian Institute of Health and Welfare
B cohort
the LSAC baby cohort comprising children who were born between March
2003 and February 2004
BIC
Bayesian information criterion
BMI
body mass index
CFI
comparative fit index
K cohort
the LSAC kindergarten cohort comprising children who were born between
March 1999 and February 2000
LSAC
Longitudinal Study of Australian Children
NAPLAN
National Assessment Program – Literacy and Numeracy
NTER
Northern Territory Emergency Response
OLS
Ordinary Least Squares regression
P1
parent 1
P2
parent 2
PLE
parent living elsewhere
RMSEA
root mean square error of approximation
RRR
relative risk ratio
SDQ
Strengths and Difficulties Questionnaire
SEIFA
Socio-Economic Indexes for Areas— a suite of four indexes that have been
created from social and economic Census information
SEP
socioeconomic position
SRMR
standardised root mean square residual
VLMR
Vuong-Lo-Mendell-Rubin
77
References
Adema, W 2012, ‘Greater gender equality: what consequences for socioeconomic participation of
families?’, keynote address at the 12th Australian Institute of Family Studies Conference
Family transitions and trajectories, Melbourne, 25–27 July.
Alati, R, Maloney, E, Hutchinson, DM, Najman, JK, Mattick, RP, Bor, W & Williams, GM 2010, ‘Do
maternal parenting practices predict problematic patterns of adolescent alcohol
consumption?’, Addiction, vol. 105, no. 5, pp. 872–80.
Allison, P 1990, ‘Change scores as dependent variables in regression analysis’, Sociological
Methodology, vol. 20, pp. 93–114.
Australian Institute of Family Studies 2012, AIFS research directions 2012–15: Australian families in
a rapidly changing world, Australian Institute of Family Studies, Melbourne.
Australian Institute of Health and Welfare 2012a, Adoptions Australia 2010–11, Child Welfare Series
no. 52, Australian Institute of Health and Welfare, Canberra.
——2012b, Social and emotional wellbeing: development of a children’s headline indicator, Cat. no.
PHE 158, Australian Institute of Health and Welfare, Canberra.
Azar, ST 2002, ‘Parenting and child maltreatment’, in MH Bornstein (ed.), Handbook of parenting:
practical issues in parenting, vol. 5, 2nd ed., Laurence Erlbaum Associates, Mahwah, pp.
361–88.
Babington, B 2011, ‘National Framework for Protecting Australia’s Children: perspectives on
progress and challenges’, Family Matters, no. 89, pp. 11–20.
Baumrind, D & Black, AE 1967, ‘Socialization practices associated with dimensions of competence in
preschool boys and girls’, Child Development, vol. 38, no. 2, pp. 291–327.
Belsky, J 1984, ‘The determinant of parenting: a process model’, Child Development, vol. 55, no. 1,
pp. 83–96.
Bowes, J, Watson, J & Pearson, E 2009, ‘Families as a context for children’, in J Bowes & R Grace
(eds), Children, families and communities: contexts and consequences, Oxford University
Press, Oxford, pp. 91–110.
Bowlby, L 1988, A secure base: clinical applications of attachment theory, Tavistock/Routledge,
London.
Broderick, CB 1993, Understanding family process: basics of family systems theory, Sage, Newbury
Park.
Bromfield, L, Arney, F & Higgins, D 2014, ‘Child protection: legal grounds for intervening in family
life to ensure child safety’, in A Hayes & D Higgins (eds), Families, policy and the law: selected
essays on contemporary issues for Australia, Australian Institute of Family Studies, Melbourne,
pp. 121–29.
Bromfield, LM, Lamont, A, Parker, R & Horsfall, B 2011, Issues for the safety and wellbeing of
children in families with multiple and complex problems: the co-occurrence of domestic
violence, parental substance misuse, and mental health problems, National Child Protection
Clearinghouse Issues Paper no. 33, Australian Institute of Family Studies, Melbourne.
Buckmaster, L & Ey, C 2012, Is income management working? Parliamentary Library Background
Note, Parliament of Australia, Canberra.
Cole, TJ, Bellizzi, MC, Flegal, KM & Dietz, WH 2000, ‘Establishing a standard definition for child
overwieght and obesity worldwide: an international survey’, British Medical Journal, vol.
320, no. 7244, pp. 124–43.
78
Cole, TJ, Flegal, KM, Nicholls, D & Jackson, AA 2007, ‘Body mass index cut offs to define thinness
in children and adolescents: international survey’, British Medical Journal, vol. 335, no. 7612,
pp. 194–201.
Council of Australian Governments (COAG) 2009a, Protecting children is everyone’s business:
National Framework for Protecting Australia’s Children 2009–2020, Department of Families,
Housing, Community Services and Indigenous Affairs, Canberra.
——2009b, National plan to reduce violence against women and their children, Department of
Families, Housing, Community Services and Indigenous Affairs, Canberra.
Cox, MJ & Paley, B 1997, ‘Families as systems’, Annual Review of Psychology, vol. 48, pp. 243–67.
Croucher, R 2012, ‘Family law: challenges for responding to family violence in a federal system’, in
A Hayes & D Higgins (eds), Families, policy and the law: selected essays on contemporary
issues for Australia, Australian Institute of Family Studies, Melbourne , pp. 207–14.
Cummings, ME 1987, ‘Coping with background anger in early childhood’, Child Development, vol.
58, no. 4, pp. 976–84.
Cummings, ME, Ballard, M, El-Sheik, M & Lake, M 1991, ‘Resolution and children’s responses to
interadult anger’, Developmental Psychology, vol. 27, no. 3, pp. 462–70.
Cummings, ME, Zahn-Waxler, C & Radke-Yarrow, M 1981, ‘Young children’s responses to
expressions of anger and affection by others in the family’, Child Development, vol. 52, pp.
1274–82.
Cummings, ME & Davies, PT 2002, ‘The effects of marital conflict on children: recent advances and
emerging themes in process-oriented research’, Journal of Child Psychology and Psychiatry,
vol. 43, no. 1, pp. 31–63.
—— 2010, Marital conflict and children: an emotional security perspective, The Guildford Press,
New York.
Cummings, ME, Goeke-Morey, M & Papp, LM 2003, ‘Children’s responses to everyday marital
conflict tactics in the home’, Child Development, vol. 74, no. 6, pp. 1918–29.
Davies, PT, Cummings, ME & Winter, MA 2004, ‘Pathways between profiles of family functioning,
child security in the interparental subsystem, and child psychological problems’, Development
and Psychopathology, vol. 16, no. 3, pp. 252–550.
Davies, PT, Sturge-Apple, ML, Cicchetti, D & Cummings, ME 2008, ‘Adrenocortical underpinnings
of children’s psychological reactivity to interparental conflict’, Child Development, vol. 79,
no. 6, pp. 1693–706.
Gaudin, JM, Polansky, NA, Kilpatrick, AC & Shilton, P 1996, ‘Family functioning in neglectful
families’, Child Abuse and Neglect, vol. 20, no. 4, pp. 363–77.
Golombok, S 2000, Parenting: what really counts? Routledge, London.
Goodman, R 1997, ‘The Strengths and Difficulties Questionnaire: a research note’, Journal of Child
Psychology and Psychiatry, vol. 38, no. 5, pp. 581–6.
Gray, M & Smart, D 2009, ‘Growing Up in Australia: The Longitudinal Study of Australian Children:
a valuable new data source for economists’, Australian Economic Review, vol. 42, no. 3, pp.
367–76.
Hamilton, M & Redmond, G 2010, Conceptualisation of social and emotional wellbeing for children
and young people, and policy implication, Australian Research Alliance for Children and
Youth/Australian Institute of Health and Welfare, Canberra.
Hand, K, Gray, M, Higgins, D, Lohoar, S & Deblaquiere, J 2011, Life around here: community, work
and family life in three Australian communities (AIFS Research Report no. 19), Australian
Institute of Family Studies, Melbourne.
79
Hetherington, ME 2006, ‘The influence of conflict, marital problem solving and parenting on
children’s adjustments in nondivorced, divorced and remarried families’, in A Clarke-Stewart
& J Dunn (eds), Families count: effects on child and adolescent development, Cambridge
University Press, Cambridge.
Higgins, D & Katz, I 2008, ‘Enhancing service systems for protecting children: promoting child
wellbeing and child protection reform in Australia’, Family Matters, vol. 80, pp. 43–50.
Higgins, D J 2007, Cooperation and coordination: an evaluation of the Family Court of Australia’s
Magellan case-management model, Family Court of Australia, Canberra.
——2010a, ‘Community development approaches to safety and wellbeing of Indigenous children’,
Closing the Gap Clearinghouse Resource Sheet no. 1, Closing the Gap Clearinghouse,
Australian Institute of Health and Welfare/Australian Institute of Family Studies, Canberra.
——2010b, ‘Impact of past adoption practices: summary of key issues from Australian research’,
Australian Journal of Adoption, vol. 2, no. 2.
——2011, ‘Protecting children: evolving systems’, Family Matters, vol. 89, pp. 5–10.
Higgins, DJ & McCabe, MP 1998, ‘Parent perceptions of maltreatment and adjustment in children’,
Journal of Family Studies, vol. 4, pp. 53–76.
Higgins, DJ & McCabe, MP 2000, ‘Multi-type maltreatment and the long-term adjustment of adults’,
Child Abuse Review, vol. 9, pp. 6–18.
Holden, GW 2010, Parenting: a dynamic perspective, Sage, Thousand Oaks.
Holmes-Smith, P 2012, Structural equation modelling (using Mplus): from the fundamental to
advanced topics, School Research Evaluation and Measurement Services (SREAMS),
Melbourne.
Holzer, PJ, Higgins, J, Bromfield, LM, Richardson, N & Higgins, D J 2006, The effectiveness of
parent education and home visiting child maltreatment prevention programs, Child Abuse
Prevention Issues no. 24, National Child Protection Clearinghouse, Australian Institute of
Family Studies, Melbourne.
Horsfall, B, Bromfield, L & McDonald, M 2010 Are social marketing campaigns effective in
preventing child abuse and neglect?, National Child Protection Clearinghouse Issues, no. 32,
Australian Institute of Family Studies, Melbourne.
Howe, B 2012, ‘Lives on hold: unlocking the potential of Australia’s workforce’ seminar presentation
at the Australian Institute of Family Studies, 14 August.
Hunter, C 2011, Defining the public health model for the child welfare services context, National
Child Protection Clearinghouse Resource Sheet, Australian Institute of Family Studies,
Melbourne.
Hunter, C & Price-Robertson, R 2012, Family structure and child maltreatment: do some family types
place child at greater risk? Child Family Community Australia Paper no. 10, Australian
Institute of Family Studies, Melbourne.
Johnson, D 2005, ‘Two-wave panel analysis: comparing statistical methods for studying the effects of
transitions’, Journal of Marriage and Family, vol. 67, pp. 1061–75.
Johnson, VK 2003, ‘Linking changes in whole family functioning and children’s externalizing
behavior across the elementary school years’, Journal of Family Psychology, vol. 17, no. 4,
pp. 499–509.
Kerrig, PK 1995, ‘Triangles in the family circle: effects of family structure on marriage, parenting,
and child adjustment’, Journal of Family Psychology, vol. 9, no. 1, pp. 28–43.
Krishnakumar, A & Buehler, C 2000, ‘Interparental conflict and parenting behavior: a meta-analytic
review’, Family Relations, vol. 49, no. 1, pp. 25–44.
The Lancet 2010, ‘Health inequality—an election manifesto?’ (editorial) vol. 375, no. 9714, p. 525.
80
Lo, Y, Mendell, NR & Rubin, DB 2001, ‘Testing the number of components in a normal mixture’,
Biometrika, vol. 88, pp. 767–78.
Lucas, N, Nicholson, JM & Maguire, B 2011, ‘Parenting practices and behaviours’, in B Maguire
(ed.), The Longitudinal Study of Australian Children Annual Statistical Report 2010,
Australian Institute of Family Studies, Melbourne.
Maccoby, EE & Martin, JA 1983, ‘Socialization in the context of the family: parent–child interaction’,
in PH Mussen & EM Hetherington (eds), Handbook of child psychology Volume 4:
socialization, personality, and social development, John Wiley and Sons, New York.
McLanahan, S & Sandefur, G 1994, Growing up with a single parent: what hurts, what helps?
Harvard University Press, Cambridge.
Mildon, R & Polimeni, M 2012, ‘Parenting in the early years: effectiveness of parenting support
programs for Indigenous families’, Resource sheet no. 16 , Closing the Gap Clearinghouse,
Australian Institute of Health and Welfare, Canberra & Australian Institute of Family Studies,
Melbourne.
Minuchin, S 1978, Families and family therapy, Tavistock Publications, London.
Muir, K, Katz, I, Purcal, C, Patulny, R, Flaxman, S, Abello, D, Cortis, N, Thomson, C, Oprea, I, Wise,
S, Edwards, B, Gray, M & Hayes, A 2009, National evaluation (2004–2008) of the Stronger
Families and Communities Strategy 2004–2009, Occasional Paper no. 24, Department of
Families, Housing, Community Services and Indigenous Affairs, Canberra.
O’Connor, TG & Scott, SBC 2007, Parenting and outcomes for children, Joseph Rowntree
Foundation, York.
O’Donnell, M, Scott, D & Stanley, F 2008, ‘Child abuse and neglect—is it time for a public health
approach?’ Australian and New Zealand Journal of Public Health, vol. 32, no. 4, pp. 325–30.
Pettit, GS, Bates, JE & Dodge, K 1997, ‘Supportive parenting, ecological context, and children’s
adjustment: a seven-year longitudinal study’, Child Development, vol. 68, no. 5, pp. 908–23.
Pettit, GS & Bates, JE 1989, ‘Family interaction patterns and children’s behavior problems from
infancy to 4 years’, Developmental Psychology, vol. 25, no. 3, pp. 413–20.
Prinz, RJ, Sanders, MR, Shapiro, CJ, Whitaker, DJ & Lutzker, JR 2009, ‘Population-based prevention
of child maltreatment: the U.S. Triple P System Population Trial’, Prevention Science, vol. 10,
no. 1, pp. 1–2.
Repetti, RL, Taylor, SE & Seeman, TE 2002, ‘Risky families: family social environments and the
mental and physical health of offspring’, Psychological Bulletin, vol. 128, no. 2, pp. 330–66.
Sanders, MR 2008, ‘Triple P-Positive Parenting Program as a public health approach to strengthening
parenting’, Journal of Family Psychology, vol. 22, no. 3, pp. 506–17.
Sanders MR 2012, ‘Development, evaluation, and multinational dissemination of the Triple P-Positive
Parenting Program’, Annual Review of Clinical Psychology, vol. 8, pp. 345–79.
Sanders, MR, Cann, W & Markie-Dadds, C 2003, ‘The Triple P-Positive Parenting Programme: a
universal population-level approach to the prevention of child abuse’, Child Abuse Review,
vol. 12, no. 3, pp. 155–71.
Sanders, MR & Pidgeon, A 2011, ‘The role of parenting programmes in the prevention of child
maltreatment’, Australian Psychologist, vol. 46, pp. 199–209.
Sanders, MR & Prinz, RJ 2008, ‘Using the mass media as a population level strategy to strengthen
parenting skills’, Journal of Clinical Child and Adolescent Psychology, vol. 37, no. 3, pp.
609–21.
Sanders, MR, Ralph, A, Thompson, R, Sofronoff, K, Gardiner, P, Bidwell, K & Dwyer, S 2005, Every
family: a public health approach to promoting children’s wellbeing, brief report, University
of Queensland, Brisbane.
81
Saunders, BJ & Goddard, C 2002, The role of mass media in facilitating community education and
child abuse prevention strategies, Child Abuse Prevention Issues no. 16, Australian Institute
of Family Studies, Melbourne.
Schoppe-Sullivan, SJ, Schermerhorn, AC & Cummings, ME 2007, ‘Marital conflict and children’s
adjustment: evaluation of the parenting process model’, Journal of Marriage and Family, vol.
69, no. 5, pp. 1118–34.
Scott, D 2006, ‘Towards a public health model of child protection in Australia’, Communities,
Children and Families in Australia, vol. 1, no. 1, pp. 9–16.
Scott, D & Higgins, D 2011, ‘Supporting families’ (Chapter 7), in Northern Territory Emergency
Response Evaluation Report 2011, Department of Families, Housing, Community Services
and Indigenous Affairs, Canberra.
Scott, D, Higgins, D & Franklin, R 2012, The role of supervisory neglect in childhood injury, Child
Family Community Australia paper no. 08, Australian Institute of Family Studies, Melbourne.
Senechal, M & Schagen, I 2002, ‘Parental involvement in the development of children’s reading skill:
a five-year longitudinal study’, Child Development, vol. 73, no. 2, pp. 445–60.
Sivak, L, Arney, F & Lewig, K 2008, A pilot exploration of a family home visiting program for
families of Aboriginal and Torres Strait Islander children, report and recommendations:
perspectives of parents of Aboriginal children and organisational considerations,
Australian Centre for Child Protection, Adelaide, retrieved 6 February 2012.
Soloff, C, Lawrence, D & Johnstone, R 2005, LSAC sample design, LSAC Technical Paper no. 1,
Australian Institute of Family Studies, Melbourne.
StataCorp 2011, Stata user’s guide: release 12, Stata Press, College Station.
Sturge-Apple, ML, Davies, PT & Cummings, ME 2010, ‘Typologies of family functioning and
children’s adjustment during the early school years’, Child Development, vol. 81, no. 4, pp.
1320–35.
Sweeting, HN 2007, ‘Measurement and definitions of obesity in childhood and adolescence: a field
guide for the uninitiated’, Nutrition Journal, vol. 6, pp. 32–39.
Tamis-LeMonda, CS, Užgiris, IC & Bornstein, MH 2002, ‘Play in parent–child interactions’, in MH
Bornstein (ed.), Handbook of Parenting: Practical Issues in Parenting, Volume 5, 2nd ed.,
Laurence Erlbaum Associates, Mahwah.
Taylor, D, Stanton, D & Gray, M 2012, ‘New conditionality in Australian social security policy and
child welfare’, paper presented at the 12th Australian Institute of Family Studies Conference:
Family transitions and trajectories, Melbourne, 25–27 July.
Vermunt, JK & Magison, J 2002, ‘Latent class cluster analysis’, in JA Hagenaars & AL McCutcheon
(eds), Applied Latent Class Analysis, Cambridge University Press, Cambridge.
Wake, M & Maguire, B 2012, ‘Children’s body mass index: cohort, age and socio-economic
influences’, in B Maguire (ed.), The Longitudinal Study of Australian Children: Annual
statistical report, Australian Institute of Family Studies, Melbourne.
Wise, S 2003, Family structure, child outcomes and environmental mediators: an overview of the
development in diverse families study, Research Paper no. 30, Australian Institute of Family
Studies, Melbourne.
Wise, S, da Silva, L, Webster, E & Sanson, A 2005, The efficacy of early childhood interventions,
AIFS research report no. 14, Australian Institute of Family Studies, Melbourne.
Zubrick, SR, Smith, GJ, Nicholson, JM, Sanson, AV, Jackiewcz, TA & LSAC Research Consortium
2008, Parenting and families in Australia , Social Policy Research Paper no. 34, Department
of Families, Housing, Community Services and Indigenous Affairs, Canberra.
82
Endnotes
National Assessment Program – Literacy and Numeracy
2 Children who have been removed from the care of their parent(s)—also referred to as ‘looked after’ children,
or children in ‘alternative care’.
3 Babington (2011) provided a brief history and context to the development of a national approach to address
child abuse and neglect in Australia.
4
A range of risk factors for child abuse and neglect were targeted, including: poverty; substance abuse; mental
illness; maternal education; overcrowding or unstable housing; health (including children with disabilities or low
birth weight); dangerous/unsafe communities; drug/alcohol misuse; unemployment; and parenting skills and
capacity (including the stresses on single parents, parents who were themselves victims of child abuse, neglect or
poor parenting).
5
CFI (comparative fit index) greater than 0.95 and SRMR (standardized root mean square residual) less than
0.05 indicate good construct validity.
6
The difference between these coefficients is non-trivial, and closer inspection revealed that only around 10 per
cent of children who were reported as with disability at 8–9 years were also reported as with disability at 10–11
years. In other words, this indicator is pointing to two largely different groups of children at these two points in
time. Note that those children who were 8–9 years with disability have significantly lower Year 5 NAPLAN
scores than those who are not reported as living with disability, as we might expect. Further research could help
us understand more about this group of children.
7
The main advantage of the fixed effects approach we adopt is that it controls for all observed and unobserved
differences between individuals that do not change over time (note that all covariates that do not change over
time are dropped from the analysis). The key disadvantage, following directly from this, is that we do not exploit
variation between individuals. This can lead to larger standard errors with fixed effects (thus a higher threshold
for detecting statistically significant results). We experimented with the random effects approach, which exploits
between subject variation (as well as within subject variation) and found our results were substantively
equivalent (results not shown). Moreover, the Hausman specification test (StataCorp 2011) clearly rejected the
use of random effects in favour of fixed effects.
1
8
For other examples of parenting programs that have been evaluated, see Holzer et al. (2006), and for a list of
other publications, see the Australian Institute of Family Studies website.
9
Other criteria relate to implementing programs.
10
For example: DrinkWise Australia’s ‘Kids Absorb your drinking’ campaign; DrinkWise Australia’s ‘Under
your influence’ campaign; NAPCAN’s ‘Children See: Children Do’ campaign.
11
See the World Health Organization website.
12
Fair society, healthy lives: the Marmot review
83